WO2016155537A1

WO2016155537A1 - Method and device for ranking search results of picture objects

Info

Publication number: WO2016155537A1
Application number: PCT/CN2016/076951
Authority: WO
Inventors: 黄恒
Original assignee: 阿里巴巴集团控股有限公司
Priority date: 2015-03-30
Filing date: 2016-03-22
Publication date: 2016-10-06
Also published as: CN106156063B; CN106156063A

Abstract

Provided in the present invention are a method and device for ranking search results of picture objects. The method for ranking search results of picture objects comprises: in response to receiving a search request of a user, obtaining picture objects matching the search request; extracting visual characteristics of the obtained picture objects; obtaining comprehensive preference scores of the user to the obtained picture objects on the basis of preference scores of the user to the visual characteristics; ranking the picture objects according to the comprehensive preference scores of the user to the obtained picture objects. The present invention enables ranking of the search results of the picture objects to be more accurate and close to the user interest.

Description

Related method and device for sorting picture object search results

The present application claims the priority of the Chinese Patent Application No. 201510144467.8, filed on March 30, 2015, the disclosure of which is incorporated herein by reference. in.

Technical field

The present invention relates to the field of Internet technologies, and in particular, to a related method and apparatus for ordering picture object search results.

Background technique

With the development of Internet technology, the amount of information on the Internet is also increasing. When users search for the information they need through search engines, they often cannot accurately describe the information they need or match the search keywords of the user. Due to the large amount of information, the search engine returns a large amount of relevant information as a search result. The more information the search result contains, the more difficult it is for the user to select the information he needs in the search results.

In order to solve the above problem, some search engines provide a scheme for sorting search results based on user interest, specifically: clustering based on text features of user search results, thereby determining information of interest to the user, when the user subsequently performs in the search engine. The search results are sorted based on information of interest to the user during the search. Wherein, if a text feature content of a search result is large, an information extraction operation is required, since it is difficult to quantize at the time of information extraction, and if the text feature is not a regular form type, it is difficult to extract the feature therein. At the same time, different text descriptions may be used for the same attribute in different search results. For example, the attribute silk and silk of the product are actually the same attribute. Therefore, the search result needs to be normalized, if the normalization is not processed. If you are accurate, you cannot identify the two identical attributes to the same point of interest. In addition, since the same keyword has different meanings under different classifications, for example, for the keyword apple, the meanings represented by different categories such as fruit, movie, and 3C are different, so it is also necessary to calculate the semantic scene according to the context information. Disambiguating the keyword, if the context information cannot be obtained, the disambiguation result will be affected, and the search result obtained by searching according to the disambiguated keyword will contain a large amount of information that does not match the real needs of the user, based on This information is clustered and it will be difficult to accurately express the user's true point of interest.

It can be seen that determining the information of interest to the user based on the text feature requires information extraction and normalization of the text feature. And complicated technical processing such as elimination, it is not only difficult to extract attribute information (that is, low extraction efficiency), and the accuracy of user interest analysis based on the attribute information extracted based on text features is low, and it is difficult to accurately express the user's true interest. Points, based on the sorting of search results, users are also more difficult to find the information they really need.

Summary of the invention

One of the technical problems solved by the present invention is to provide a related method and apparatus for sorting search results of picture objects, which can make the ranking of search results of picture objects more accurate and close to user interests.

According to an embodiment of an aspect of the present invention, a method for assisting in sorting picture object search results is provided, including:

Obtaining the image object involved in the user behavior record within the set time range;

Extracting visual features of the acquired picture object;

Determining the number of picture objects having the same visual feature in the acquired picture object;

Based on the number of picture objects having the same visual feature, the user's preference score for the visual feature is determined for ordering the picture object search results.

Optionally, the visual feature comprises at least one of the following:

Picture object brightness, picture object saturation, picture object sharpness, picture object contrast, picture object brilliance, picture object puzzle, picture object body area ratio, picture object body color depth, picture object body color scheme.

Optionally, determining a user's preference score for the visual feature based on the number of picture objects having the same visual feature comprises:

Determining a base score of the visual feature based on a number of picture objects having the same visual feature and a correction value, wherein the correction value represents a ratio of the picture object having the visual feature in the picture object library;

A user's preference score for the visual feature is determined based on the base score of the visual feature.

Optionally, the behavior record comprises a selection.

Optionally, the behavior record comprises: browsing and selecting, and determining a user's preference score for the visual feature based on the number of picture objects having the same visual feature comprises:

Determining a base score of the visual feature based on a number of picture objects having the same visual feature and a corresponding weighting factor, wherein the weighting factor is determined based on whether the behavior record is browsed or selected;

Optionally, the visual features are characterized by precise visual feature values.

Optionally, the visual feature is characterized by a precise visual feature value or a hierarchical visual feature value, wherein in the case that the visual feature is determined by the hierarchical visual feature value standard, extracting the visual feature of the acquired image object comprises:

Extracting precise visual feature values of the visual features of the acquired picture object;

A hierarchical visual feature value to which the precise visual feature value belongs is determined as a visual feature of the acquired picture object.

According to an embodiment of another aspect of the present invention, a method for ordering picture object search results is provided, including:

Acquiring a picture object that matches the search request in response to receiving a search request of the user;

Extracting visual features of the acquired picture object;

Obtaining a comprehensive preference score of the user for the acquired picture object based on the user's preference score for the visual feature, wherein the user's preference score for the visual feature is determined according to the method for sorting the search results for the auxiliary picture object. of;

The picture objects are sorted according to a user's comprehensive preference score for the acquired picture object.

Optionally, the visual feature comprises at least one of the following:

Optionally, obtaining a comprehensive preference score of the user for the acquired picture object based on the user's preference score for the visual feature includes:

The sum of the user's preference scores for the visual features is taken as the user's overall preference score for the acquired picture object.

Optionally, acquiring the picture object that matches the search request further includes: acquiring an original sort score of the picture object that matches the search request, and obtaining the user pair obtained based on the user's preference score for the visual feature. The composite preference score for a picture object includes:

Adding the sum of the preference scores of the user to the visual feature to the obtained original sort score of the image object as a comprehensive preference score of the user for the acquired image object; or

The sum value obtained by adding the sum of the preference scores of the visual features to the predetermined adjustment value plus the obtained original sort score of the picture object is used as a comprehensive preference score of the user for the acquired picture object.

Optionally, sorting the picture objects according to a user's comprehensive preference score of the acquired picture object includes:

The picture object row is ranked according to the user's comprehensive preference score of the acquired picture object from high to low. sequence.

According to an embodiment of the present invention, there is provided an apparatus for assisting in sorting picture object search results, comprising:

An obtaining unit, configured to acquire a picture object involved in a user behavior record in a set time range;

An extracting unit, configured to extract a visual feature of the acquired picture object;

a number determining unit, configured to determine a number of picture objects having the same visual feature among the acquired picture objects;

And a preference score determining unit, configured to determine a user's preference score for the visual feature based on the number of picture objects having the same visual feature, for ordering the picture object search result.

Optionally, the visual feature comprises at least one of the following:

Optionally, the preference score determining unit comprises:

a first base score determining subunit for determining a base score of the visual feature based on a number of picture objects having the same visual feature and a correction value, wherein the correction value indicates a ratio of the picture object having the visual feature in the picture object library ;

The first preference score determining subunit is configured to determine a user's preference score for the visual feature based on the base score of the visual feature.

Optionally, the behavior record comprises a selection.

Optionally, the behavior record includes: browsing and selecting, and the preference score determining unit comprises:

a second base score determining subunit, configured to determine a base score of the visual feature based on a number of picture objects having the same visual feature and a corresponding weighting factor, wherein the weighting factor is determined according to whether the behavior record is browsed or selected;

A second preference score determining subunit is configured to determine a user's preference score for the visual feature based on the base score of the visual feature.

Optionally, the visual feature is characterized by an accurate visual feature value or a hierarchical visual feature value, wherein in the case that the visual feature is represented by the hierarchical visual feature value, the extracting unit comprises:

Extracting a subunit for extracting an accurate visual feature value of a visual feature of the acquired picture object;

a hierarchical sub-unit for determining a hierarchical visual feature value to which the accurate visual feature value belongs, as obtained The visual characteristics of the taken picture object.

According to an embodiment of still another aspect of the present invention, an apparatus for ordering picture object search results is provided, including:

a picture object obtaining unit, configured to acquire a picture object that matches the search request in response to receiving a search request of the user;

a visual feature extraction unit, configured to extract a visual feature of the acquired picture object;

a comprehensive preference score obtaining unit, configured to obtain a user's comprehensive preference score for the acquired picture object based on a user's preference score for the visual feature, wherein the user's preference score for the visual feature is used by the above Determining the device for sorting the search results of the auxiliary picture object;

a sorting unit, configured to sort the picture objects according to a user's comprehensive preference score of the acquired picture objects.

Optionally, the visual feature comprises at least one of the following:

Optionally, the integrated preference score obtaining unit is used to:

Optionally, the picture object obtaining unit is further configured to: obtain an original sort score of the picture object that matches the search request, where the integrated preference score obtaining unit is used to:

The sum of the user's preference scores of the visual features and the predetermined adjustment value plus the obtained original sort score of the picture object is used as the comprehensive preference score of the user for the acquired image object.

Optionally, the sorting unit is used to:

The picture objects are sorted in descending order of the user's overall preference scores for the acquired picture objects.

The embodiment of the present application is based on the number of picture objects having the same visual feature in the picture object involved in the user behavior record, and determining the user's preference for the visual feature, and using the preference to sort the search result of the picture object, Sorting the search results of image objects is more accurate and close to user interest.

Since the picture object can express the characteristics of the object more intuitively than the text description, the view based on the picture object The user's preference score for the visual feature can be more accurately and closely related to the user's interest, while the visual feature of the picture object is easier to extract the attribute of the object from the text description.

Those skilled in the art will appreciate that although the following detailed description is made with reference to the illustrated embodiments and drawings, the invention is not limited to these embodiments. Rather, the scope of the invention is intended to be limited the scope of the invention

DRAWINGS

Other features, objects, and advantages of the invention will be apparent from the description of the appended claims. The same or similar reference numerals in the drawings denote the same or similar components. 1 is a flow chart of a method for assisting in sorting picture object search results, in accordance with one embodiment of the present invention;

2 is a flowchart of a method of determining a user's preference score for the visual feature based on the number of picture objects having the same visual feature, in accordance with one embodiment of the present invention;

3 is a schematic diagram of a method for determining a color depth correction value of a picture object body according to an embodiment of the present invention;

4 is a schematic diagram of determining a score of a color depth of a subject of a picture object according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of the number of picture objects of each visual feature according to an embodiment of the present invention; FIG.

6-1 is a schematic diagram of preference scores of picture object brightness according to an embodiment of the present invention;

6-2 is a schematic diagram of preference scores of picture object sharpness according to an embodiment of the present invention;

7 is a flowchart of a method of determining a user's preference score for the visual feature based on the number of picture objects having the same visual feature, in accordance with another embodiment of the present invention;

8 is a flow chart of a method for ordering picture object search results, in accordance with one embodiment of the present invention;

9-1 is a schematic diagram showing an original sorting of search results according to an embodiment of the present invention;

9-2 is a schematic diagram of sorting search results according to a composite preference score, in accordance with one embodiment of the present invention;

FIG. 10 is a schematic structural diagram of an apparatus for assisting in sorting picture object search results according to an embodiment of the present invention; FIG.

11 is a schematic structural view of an extraction unit according to an embodiment of the present invention;

FIG. 12 is a schematic structural diagram of a preference score determining unit according to an embodiment of the present invention; FIG.

FIG. 13 is a schematic structural diagram of a preference score determining unit according to another embodiment of the present invention; FIG.

FIG. 14 is a schematic structural diagram of an apparatus for sorting picture object search results according to an embodiment of the present invention.

detailed description

Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as a process or method depicted as a flowchart. Although the flowcharts describe various operations as a sequential process, many of the operations can be implemented in parallel, concurrently or concurrently. In addition, the order of operations can be rearranged. The process may be terminated when its operation is completed, but may also have additional steps not included in the figures. The processing may correspond to methods, functions, procedures, subroutines, subroutines, and the like.

The computer device includes a user device and a network device. The user equipment includes, but is not limited to, a computer, a smart phone, a PDA, etc.; the network device includes but is not limited to a single network server, a server group composed of multiple network servers, or a cloud computing based computer Or a cloud composed of a network server, wherein cloud computing is a type of distributed computing, a super virtual computer composed of a group of loosely coupled computers. Wherein, the computer device can be operated separately to implement the present invention, and can also access the network and implement the present invention by interacting with other computer devices in the network. The network in which the computer device is located includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a VPN network, and the like.

It should be noted that the user equipment, the network equipment, the network, and the like are merely examples, and other existing or future possible computer equipment or networks, such as those applicable to the present invention, are also included in the scope of the present invention. It is included here by reference.

The methods discussed below, some of which are illustrated by flowcharts, can be implemented in hardware, software, firmware, middleware, microcode, hardware description languages, or any combination thereof. When implemented in software, firmware, middleware or microcode, the program code or code segments to carry out the necessary tasks can be stored in a machine or computer readable medium, such as a storage medium. The processor(s) can perform the necessary tasks.

The specific structural and functional details disclosed are merely representative and are for the purpose of describing exemplary embodiments of the invention. The present invention may, however, be embodied in many alternative forms and should not be construed as being limited only to the embodiments set forth herein.

It should be understood that although the terms "first," "second," etc. may be used herein to describe the various elements, these elements should not be limited by these terms. These terms are used only to distinguish one unit from another. For example, a first unit could be termed a second unit, and similarly a second unit could be termed a first unit, without departing from the scope of the exemplary embodiments. The term "and/or" used herein includes any and all combinations of one or more of the associated listed items.

It will be understood that when a unit is referred to as "connected" or "coupled" to another unit, it can be directly connected or coupled to the other unit, or an intermediate unit can be present. In contrast, when a unit is referred to as being "directly connected" or "directly coupled" to another unit, there is no intermediate unit. Should be solved in a similar way The other words used to describe the relationship between the units (eg "between" and "directly between", "close to" compared to "with" Directly adjacent to "etc."

The terminology used herein is for the purpose of describing the particular embodiments, The singular forms "a", "an", It is also to be understood that the terms "comprising" and """ Other features, integers, steps, operations, units, components, and/or combinations thereof.

It should also be noted that, in some alternative implementations, the functions/acts noted may occur in a different order than that illustrated in the drawings. For example, two figures shown in succession may in fact be executed substantially concurrently or sometimes in the reverse order, depending on the function/acts involved.

The inventors have found that when a user searches for a picture object, it often exhibits a preference for a picture object having a certain type of visual feature, for example, some users are particularly interested in a picture object having a higher brightness of the picture object. Therefore, the embodiment of the present application extracts the visual feature of the picture object, determines the user's preference score for the visual feature, and thus can use the preference score to sort the search result of the picture object, so that the search result of the picture object can be more accurately and closely matched to the user. interest.

The invention is further described in detail below with reference to the accompanying drawings.

1 is a flow chart of a method for assisting in sorting picture object search results according to an embodiment of the present invention. As shown in FIG. 1, the method mainly includes the following steps:

S10. Acquire a picture object involved in a user behavior record in a set time range;

S11. Extract a visual feature of the acquired picture object.

S12. Determine a number of picture objects having the same visual feature in the acquired picture object.

S13. Determine a user's preference score for the visual feature based on the number of picture objects having the same visual feature, and use the picture object search result to be sorted.

The above steps are further described in detail below.

The method for assisting the sorting of the picture object search results shown in FIG. 1 is not performed when the user searches for the picture object, but is performed in the server before the user searches for the picture object. It can be set to be executed once for each user on a regular basis (eg, one day, one month, etc.). In this way, changes in user preferences can be continuously tracked based on the history of the update, making the ranking more reflective of the user's changing preferences.

First, it should be noted that the final purpose of the method for assisting the sorting of the search results of the picture object in the embodiment is to determine the user's preference score for the visual feature of the picture object, so that the subsequent user searches for the picture object. When the time is selected, the picture object search result may be sorted according to the determined preference score of the user's visual features of the picture object. In the embodiment of the present application, a picture object related to a user behavior record in a set time range is used as source data, and feature data is extracted from the source data to determine a user's preference score for a visual feature of the picture object.

The set time range described in step S10 may be one month from the current time, or one year from the current time, or 10 years from the current time or other time range, etc., the setting The determination of the time range needs to ensure that sufficient source data is available within the set time range to determine the user's preference score for the visual feature. For example, the current time is December 31, 2014, and the whole month of December 2014 can be used as the set time range. The whole year of 2014 can be used as the set time range, or the tenth of 2005-2014 can be used. Year is used as the set time range. It can be understood that the more the source data acquired in the set time range, the more accurate the user's preference score for the visual feature is determined. Therefore, the “sufficient source data” is not specifically limited herein.

In an embodiment of the present application, the behavior record includes browsing and/or selection. The browsing includes but is not limited to: the user clicks to view the picture object or clicks to view the product corresponding to the picture object, and the like. The selection includes, but is not limited to, the user purchasing the product corresponding to the picture object, or the user downloading the picture object, or the user collecting the picture object, or setting the product corresponding to the picture object to the product of interest, and the like.

As can be seen from the above definition of selection and browsing, the user's interest in the selected picture object is higher than the user's interest in the browsed picture object.

In addition, it should be noted that the picture object described in this embodiment is a display picture displayed on the search result list page as a representative image of a search result, instead of appearing after clicking a search result on the search result list page. Details image in the details page. For example, for a top, the detailed picture of the top is included, and when the picture of the search result of the top is displayed, only one representative figure is displayed on the search result list page. We refer to this representative image as the picture object or display image of the search result of the top. Certainly, the embodiment of the present application does not exclude the case where the display picture is a dynamically changing picture, and the dynamic change means that there may be multiple corresponding display picture objects for the same item or information. For example, for a piece of clothing, the search result list page may have a picture when it is hung, an effect picture of the person wearing it, and a contrast picture of different colors, etc., for the case of the plurality of display pictures, the embodiment All display images can be obtained as the acquired image object.

The method for obtaining the picture object involved in the user behavior record in the set time range is as follows: the user ID (identification) searches for historical data of all behaviors of the user on the website where the server is located. The picture object selected by the user can be identified based on the first type of record in the historical data. The first type of record is, for example, a transaction record of a user recorded by the server on the website, and a download record. For example, it is recognized according to the transaction record whether the user purchases a product corresponding to the picture object, or whether the user downloads the picture object or the like according to the download record. According to the second in the historical data The class record identifies the picture object that the user is browsing. The second type of record, for example, the ipv (the number of times of browsing of the product details page) of the user who browsed the search result on the website, can identify whether the user browses the picture object and the number of times of browsing according to the ipv. Through the above method, the picture object and the corresponding number browsed by the user and the picture object selected by the user and the corresponding number can be determined.

The visual feature of the acquired picture object is extracted in step S11, wherein the visual feature comprises at least one of the following:

It can be understood that the visual features of the picture object, ie the image features, the visual features of the picture object can be characterized by corresponding visual feature values.

A method for extracting the above visual features in an embodiment of the present application includes:

The picture object brightness can be obtained by calculating the average value of the picture object in the HSV (Hue tone, Saturation saturation, Value brightness) space. The specific calculation method can be implemented by any existing technology. Make specific restrictions.

The saturation of the picture object can be obtained by calculating the average value of the S of the picture object in the HSV space. The specific calculation method can be implemented by any existing technology, and the embodiment of the present application does not specifically limit this.

Image object sharpness, the Laplacian operator template can be used to convolute the image object, and the average value obtained after the convolution calculation is the image object sharpness. The specific calculation method can be implemented by any existing technology. The application examples do not specifically limit this.

Image object glamour: can be obtained by calculating the weighted average of the image object in the RGB (red R, green G, blue B) space, pixel R / G / B component and the mean difference, the specific calculation method can be used A technical implementation is not specifically limited in this embodiment of the present application.

The image object contrast ratio can be obtained by calculating the average of the sum of the brightness of each pixel (V of the HSV space) and the brightness of the picture object. The specific calculation method can be implemented by any existing technique. This is not a specific limitation.

Whether the picture object is a jigsaw: can be identified by the continuity of the picture and the fault. The specific identification method can be implemented by any existing technology, and the embodiment of the present application does not specifically limit this. If the picture object is identified as a puzzle, the corresponding value is a first specified value, for example, 1 or 101, and otherwise is a second specified value, for example, 0 or 102, the first specified value and the second specified The value can be set as needed.

The ratio of the main area of the picture object: Since a picture object may contain multiple entities, first determine the body in the picture object (ie, the entity that the picture object focuses on), so as to surround the area of the smallest rectangular bounding box of each entity. As a determining means, the entity surrounded by the rectangular bounding box having the largest area is the main body of the image object, and the ratio of the area of the rectangular bounding frame having the largest area to the area of the entire picture object is calculated as the ratio of the main body area of the picture object.

The color of the main body of the picture object: by calculating the gray value of the main body in the picture object, generally in the range of 0-255, if the calculated gray value is greater than the specified gray threshold, it is determined that the color of the main body of the picture object is light, if less than Equal to the specified gray threshold, it is determined that the picture object body color is dark, wherein the specified gray level threshold may be, for example, 127.

Picture object main color scheme: The color of the HSV space can be normalized to the following categories: red, yellow-red, yellow, green-yellow, green, blue-green, blue, purple-blue, purple, red-violet, White, black, light gray, dark gray, to identify the color composition of the subject in the picture object (that is, which kind of color in the color classification of the main body) and the ratio of each color, the specific calculation method can be used The technical implementation of the present application does not specifically limit this. Wherein, if the ratio of only one color is greater than the specified threshold, it is determined to be a solid color, and if the ratio of the colors in 2 to 3 is greater than the specified threshold, it is determined to be a simple color; if the ratio of the three or more colors is greater than The specified threshold is then determined to be a complex suit, wherein the specified threshold may be, for example, 0.66.

Through this step S11, each visual feature of each picture object involved in the user behavior record can be acquired.

The visual feature of the picture object extracted by one of the embodiments is represented by the precise visual feature value, and the accurate visual feature value described herein is directly calculated by the method for extracting the visual feature introduced above, of course, In this embodiment, the visual feature value obtained by processing the calculated visual feature value by using the corresponding rule to remove the decimal point of the visual feature value or the specified bit of the visual feature value is not excluded as the accurate visual feature value.

The visual features of the extracted picture objects of another embodiment may be characterized by precise visual feature values or hierarchical visual feature values. The concept of the precise visual feature value is as described in the above embodiment. The hierarchical visual feature values described herein are hierarchically dividing the precise visual feature values, and each precise visual feature value is divided into one level. For example, one division method is: the 20 levels corresponding to the brightness of the picture object are respectively 1 to 20 levels; the 20 levels corresponding to the sharpness of the picture object are 21st to 40th, the 20 levels corresponding to the saturation of the picture object are the 41st to 60th levels, and the 20 levels corresponding to the picture object glare are the 61st to the 61st. At the 80th level, the 20 levels corresponding to the contrast of the picture object are the 81st to the 100th level, and the level corresponding to the picture object is the 101st to 102th level, and the ratio of the main area of the picture object corresponds to the 20th level of the 103th to the 122th level. The picture object body color scheme corresponds to the level of the 123th to 125th level, the picture object master The level corresponding to the depth of the body color is 126th to 127th grade, and each level corresponds to the corresponding precise visual feature value of the specified numerical range. Then, for the case where the extracted visual feature is a hierarchical visual feature value, it is necessary to first extract an accurate visual feature value of the acquired visual feature of the picture object, and then determine a hierarchical visual feature value to which the accurate visual feature value belongs, The determined hierarchical visual feature value is used as a visual feature of the acquired picture object.

Step S12 is for determining the number of picture objects having the same visual feature among the acquired picture objects. After the step S11 is performed, each visual feature of each picture object involved in the user behavior record is acquired, and the number of picture objects having the same visual feature in the acquired picture object can be determined. Specifically, the number of picture objects browsed by the user having the same visual feature and the number of picture objects selected by the user having the same visual feature may be determined. It can be seen that the number of picture objects having the same visual feature in the acquired picture object is determined, that is, the number of picture objects corresponding to the same visual feature involved in the user behavior record is determined.

Step S13 is to determine the user's preference score for the visual feature based on the number of picture objects having the same visual feature determined in step S12.

An embodiment is a scenario in which the picture object related to the user behavior record acquired in step S10 is a picture object selected by the user, and the preference score of the user for the visual feature is determined based on the number of picture objects having the same visual feature. The flow chart of the method is shown in Figure 2 and includes the following sub-steps:

Sub-step 20, determining a base score of the visual feature based on the number of picture objects having the same visual feature and a correction value;

Wherein the correction value represents a ratio of a picture object having the visual feature in the picture object library. Taking the color depth of the main body of the picture object as an example, the correction value includes the ratio of the picture object whose color is dark in the picture object library in the picture object library, and the picture object whose picture object color is light color. The ratio of the image object library. The library of picture objects may be, for example, a database of all picture objects that are searchable by a server on a particular website. As shown in FIG. 3, a schematic diagram of a method for determining a color depth correction value of a picture object body is shown. Suppose the picture object library contains 4 picture objects, the ID range is 1000~1003, and the picture object body color whose ID is 1000, 1001, 1003 is dark, and the picture object body color of the picture object with ID 1002 If the color is light, then the correction value of the color of the main body of the picture object in the picture object library is 0.75 (3/4, that is, the number of picture objects whose color is dark in the picture object/the number of picture objects in the picture object library) The correction value of the light color of the main body of the picture object is 0.25 (1/4, and the number of picture objects whose picture object color is light color/the number of picture objects in the picture object library). The same method can be used to determine the correction values of other visual features. It can be understood that the correction value can be calculated once at a fixed time interval, and therefore, the step of determining the correction value is not an essential step of the method.

Specifically, the base score of the visual feature is determined based on the number of picture objects having the same visual feature and the correction value, and the product of the number of picture objects having the same visual feature and the inverse of the correction value may be used as the base score of the visual feature. Taking the correction value corresponding to the color depth of the picture object body shown in FIG. 3 as an example, the color depth of the picture object body of the picture object selected by the user is as shown in the left table in FIG. 4, wherein the picture object in the picture object selected by the user The main body color is 3 dark colors, and the main body color of the picture object is 0 in a light color. The color of the main body of the picture object determined according to this is shown in the table on the right side of FIG. 4, and the color of the main body of the picture object is deep. The base score of the color is 4 (3/0.75, that is, the number of picture objects whose color is dark in the picture object/correction value), and the base color of the subject color of the picture object is 0 (0/0.25, that is, the color of the picture object body) The number of light-colored picture objects / correction value). According to this method, the basic scores of other visual features can be determined, as shown in FIG. 5, which is a schematic diagram of the number of picture objects of each visual feature (ie, the number of times each visual feature is selected by the user). According to FIG. 5, various visual features can be determined. The base score. The larger the number of picture objects of the same visual feature selected by the user (ie, the more times the user is selected), the larger the corresponding base score.

Sub-step 21, determining a user's preference score for the visual feature based on the base score of the visual feature.

After determining the basic score of each visual feature, the existing weight calculation method may be used to determine the user's preference score for the visual feature. For example, the LR (Logistic Regression) model may be used for training. This example does not impose specific restrictions.

For example, the determined preference scores of the brightness of each picture object are as shown in FIG. 6-1, and the determined preference scores of the sharpness of each picture object are as shown in FIG. 6-2, and the number of times the same visual feature is selected by the user is more. Then the value of the corresponding preference score is larger.

Another embodiment of the present application is a scenario in which the picture object related to the user behavior record acquired in step S100 includes a picture object browsed by the user and a picture object selected by the user, and determines a user pair based on the number of picture objects having the same visual feature. A flowchart of a method for the preference score of the visual feature is as shown in FIG. 7, and includes the following sub-steps:

Sub-step 70: determining a base score of the visual feature based on the number of picture objects having the same visual feature and a corresponding weighting factor;

In this embodiment, the picture object involved in the user behavior record includes a picture object browsed by the user and a picture object selected by the user, and the user's interest level with the selected picture object is higher than that of the browsed picture object, so the user can browse. The picture object and the picture object selected by the user determine different weighting factors. When determining the basic score of the visual feature, the picture object browsed by the user and the picture object selected by the user are respectively multiplied by the corresponding weighting factors.

For example, if the behavior record is a selection, you can set the weighting factor to 2. If the behavior record is browsing, you can set the weighting factor to 0.5. For example, in the set time range, the user browses the image objects whose color depth is dark in the 10 picture objects, and selects the picture objects whose color depth is dark in the five picture objects, and the color of the main body of the picture object is dark. Dark, base score = 10 × 0.5 + 5 × 2 = 15; the user in the set time range, browsed the picture object of the two picture object color shades light, selected 8 picture object body color shade For a light-colored picture object, the color of the subject of the picture object is light, and the base score is 2×0.5+8×2=17. Therefore, due to the weighting factor, although the number of times the user has recorded the behavior of the picture object whose color depth is dark (15 times) is larger than the number of times the picture object of the picture object has a light color, the picture object has a light color. (10 times), but the score is reversed.

By differently treating the picture object selected by the user and the picture object browsed by the user to determine the basic score of the visual feature, the user's interest point can be more accurately expressed.

Sub-step 71: determining a user's preference score for the visual feature based on the base score of the visual feature.

After determining the scores of the visual features, the existing weight calculation method may be used to determine the user's preference scores for the visual features. For example, the LR (Logistic Regression) model may be used for training. There are no specific restrictions on this.

In this embodiment, by acquiring the picture object involved in the user behavior record, and extracting the visual feature of the picture object, the user's preference score for the visual feature is determined based on the number of picture objects of the same visual feature, because the picture object can be more intuitively expressed. The characteristic of the product, and the visual feature of the acquired picture object is a quantized value, so the user's preference score for the visual feature determined based on the visual feature of the picture object can more accurately express the user's point of interest, and at the same time the visual of the picture object Features are easier to extract than product attributes that extract text features.

The method for assisting the sorting of the picture object search results in this embodiment can be applied to various scenes, for example, in a scene in which the search engine sorts the user's picture object search results.

The embodiment of the present application further provides a method for sorting search results of a picture object, and the flowchart of the method is as shown in FIG. 8 , and specifically includes the following steps:

S80. Acquire a picture object that matches the search request in response to receiving a search request of the user.

S81. Extract a visual feature of the acquired picture object.

S82. Obtain a comprehensive preference score of the user for the acquired picture object based on a preference score of the user for the visual feature.

S83. Sort the picture objects according to a comprehensive preference score of the acquired picture object by the user.

The method for sorting the picture object search results shown in FIG. 8 is performed in the case where the user searches for the picture object, which is determined by the user pair determined in the method for assisting the sorting of the picture object search results of FIG. The preference score of the visual feature. Since the process of FIG. 8 is performed in real time in response to a user's search request, and the process of FIG. 1 is performed periodically, for example, in the background of the server, the determined user's preference score for each visual feature may not correspond to sorting the search results. Real-time user preferences, but because the user's preferences have certain stability, periodically updating the user's preference scores for each visual feature does not affect the ranking of the search results.

The above steps are further described in detail below.

This embodiment is an application scenario of the method for assisting the sorting of the search result of the picture object in the above embodiment. For the same nouns in the embodiment, the same meanings as in the above embodiment will be omitted.

In the step S80, in response to receiving the search request of the user, the picture object that matches the search request is obtained. The method in the embodiment of the present application does not specifically limit the method for obtaining the picture object that matches the search request. The search result matching the search request may be first obtained, and then the picture object corresponding to the search result is obtained.

The visual feature in step S81 includes at least one of the following:

The method for extracting the visual features of the above visual features is the same as that described in the above embodiments, and details are not described herein again.

It should be noted that, in this embodiment, whether the visual feature of the acquired picture object is an accurate visual feature value or a hierarchical visual feature value is consistent with a visual feature obtained when determining a user's preference score for the visual feature, that is, if The visual feature acquired when the user's preference score for the visual feature is determined is the precise visual feature value. In this step, the accurate visual feature value of the visual feature of the acquired image object is also extracted, and the hierarchical visual feature value of the visual feature is acquired.

Step S82 is to obtain a comprehensive preference score of the user for the acquired picture object based on the user's preference score for the visual feature.

Wherein, the user's preference score for the visual feature can be determined by the method described in the above embodiment.

An embodiment of the present application for obtaining a comprehensive preference score of a user for an acquired picture object based on a user's preference score for the visual feature includes:

Take the visual features of the image object as only two visual features of the image object brightness and the image object sharpness, for example. The picture object brightness of a picture object belongs to the 12th level in Figure 6-1, the user's preference score for the picture object brightness level 12 is 0.02; the corresponding user's picture object sharpness for the picture object belongs to the 23rd level, The user's preference score for the picture object sharpness level 23 is 0.01, and the user's comprehensive preference score for the picture object is 0.02+0.01=0.03.

Another embodiment of the present application for obtaining a comprehensive preference score of a user for an acquired picture object based on the visual feature and a user's preference score for the visual feature includes:

Since the method of the embodiment of the present application is not used, when the website searches for a picture object that matches the search request, when the website server gives the matched picture object, the picture object is given to the picture object according to the text description of the picture object, as in the prior art. An inherent sorting, which is also often based on the sorting score of each picture object, except that the sorting score is obtained based on the degree of matching between the text description of the picture object and the keywords in the search request. The ranking score is referred to as an original ranking score, that is, the original ranking score is a ranking score assigned to the search result without considering the user's preference score for the visual feature of the embodiment. In this embodiment, the original ranking score of the picture object matching the search request is acquired while acquiring the search result matching the request; afterwards, the sum of the user's preference scores of the visual feature is added to the acquired location. The original sort score of the picture object is obtained, and the obtained sum value is used as a comprehensive preference score of the user for the acquired picture object. In this embodiment, when the search result of the picture object is sorted, the user's preference score for the visual feature and the matching degree between the search result and the search request are considered, and then the search result obtained by sorting the search result according to the integrated preference score can further improve the user experience.

Since the preference score value obtained according to the embodiment is a value less than 1, the user's preference score for the visual feature plus the original value ranking score has less influence on the original value ranking score, in order to reflect the user's vision. The effect of the preference score of the feature on the original ranking score, another embodiment of the present application may multiply the sum of the user's preference scores for the visual features by a predetermined adjustment value, and the obtained product plus the obtained original ranking score The resulting sum value is taken as a composite preference score. The predetermined adjustment value may be, for example, 100, or 10, or the like. Assuming that the original ranking score is 10 and the user's preference score for the visual feature is 0.06, then the sum of the original ranking score and the user's preference score for the visual feature is 10+0.06=10.06, if the user is visually The sum of the feature's preference scores is multiplied by a predetermined adjusted value of 100, and then added to the original sorted score is 10+0.06*100=16. It can be seen that the integrated preference score calculated by the latter has more influence on the original sorted score. Big.

Step S83: sorting the picture objects according to the user's comprehensive preference scores of the acquired picture objects includes:

The picture objects may be sorted according to a user's comprehensive preference scores of the acquired picture objects from high to low.

As shown in Figure 9-1, the original ranking of the search results is not used when the scheme of the embodiment is used, Figure 9-2 In order to use the scheme of the embodiment to sort the search results, the ranking is determined according to the size of the integrated preference score of the acquired image object by the user determined in step S820.

It can be understood that the method for ordering the picture object search results in this embodiment is performed before the search result is not displayed to the user, that is, the original ordering diagram shown in FIG. 9-1 is not used in the method of the embodiment. Will be shown to the user.

The method in this embodiment may determine a user's comprehensive preference score for the acquired picture object based on the visual feature of the search result and the user's preference score for the visual feature, and according to the user's comprehensive preference for the acquired image object. The score is the sorting of the picture object. Since the user's preference score for the visual feature determined based on the visual feature of the picture object can more accurately express the user's interest point, the sorting of the picture object based on the image object can be more convenient for the user to find the need. Search results, thereby mitigating the traffic consumption caused by users repeatedly selecting the search results they need.

The embodiment of the present application further provides an apparatus for assisting the sorting of the search results of the image objects corresponding to the method for assisting the sorting of the search results of the image objects. The structure of the apparatus is as shown in FIG. 10, and the apparatus mainly includes:

The obtaining unit 100 is configured to acquire a picture object related to the user behavior record in the set time range;

An extracting unit 101, configured to extract a visual feature of the acquired picture object;

a number determining unit 102, configured to determine a number of picture objects having the same visual feature among the acquired picture objects;

The preference score determining unit 103 is configured to determine a user's preference score for the visual feature based on the number of picture objects having the same visual feature, for the picture object search result sorting.

The set time range may be one month from the current time, or one year from the current time, or 10 years from the current time or other time range, etc., the set time range It is determined that there is a need to ensure that sufficient source data is available within the set time range to determine a user's preference score for the visual feature. For example, the current time is December 31, 2014, and the whole month of December 2014 can be used as the set time range. The whole year of 2014 can be used as the set time range, or the tenth of 2005-2014 can be used. Year is used as the set time range. It can be understood that the more the source data acquired in the set time range, the more accurate the user's preference score for the visual feature is determined. Therefore, the “sufficient source data” is not specifically limited herein.

In the embodiment of the present application, the recorded behavior includes browsing and/or selection. The browsing includes but is not limited to: the user clicks to view the picture object or clicks to view the product corresponding to the picture object, and the like. The choice includes but The user is not limited to: the user purchases the product corresponding to the picture object, or the user downloads the picture object, or the user collects the picture object, or the user sets the product corresponding to the picture object as the product of interest, and the like.

The obtaining unit 100 acquires a picture object related to the user behavior record in the set time range as follows: the user ID (identification) searches for historical data of all behaviors of the user's website at the server recorded by the server. The picture object selected by the user can be identified based on the first type of record in the historical data. The first type of record is, for example, a transaction record of a user recorded by the server on the website, and a download record. For example, it is recognized according to the transaction record whether the user purchases a product corresponding to the picture object, or whether the user downloads the picture object or the like according to the download record. The picture object browsed by the user can be identified according to the second type of record in the historical data. The second type of record, for example, the ipv (the number of times of browsing of the product details page) of the user who browsed the search result on the website, can identify whether the user browses the picture object and the number of times of browsing according to the ipv. Through the above method, the number of picture objects and corresponding numbers browsed by the user and the number of picture objects and objects selected by the user can be determined.

The visual feature of the picture object extracted by the extracting unit 101 includes at least one of the following:

Picture object brightness, picture object saturation, picture object sharpness, picture object contrast, picture object brilliance, picture object puzzle, picture object body area ratio, picture object body color depth, picture object body color scheme. The method for extracting the visual features of each visual feature by the extracting unit 101 is the same as that described in the above method embodiment, and details are not described herein again.

The visual feature of the picture object extracted by one of the embodiment extracting units 101 is represented by an accurate visual feature value table. The accurate visual feature value described herein directly calculates the obtained visual feature value by the method for extracting the visual feature introduced above. Of course, in this embodiment, the decimal point that removes the visual feature value by using the corresponding rule is not excluded, or The visual feature value obtained by processing the calculated visual feature value is obtained as a precise visual feature value by retaining a specified bit of the visual feature value or the like.

The visual feature of the picture object extracted by the extraction unit 101 of another embodiment is characterized by an accurate visual feature value or a hierarchical visual feature value, the concept of the precise visual feature value being the same as described in the above embodiment. The hierarchical visual feature values described herein are hierarchically dividing the precise visual feature values, and each precise visual feature value is divided into one level. For example, one division method is: the 20 levels corresponding to the brightness of the picture object are respectively 1 to 20 levels; the 20 levels corresponding to the sharpness of the picture object are 21st to 40th, the 20 levels corresponding to the saturation of the picture object are the 41st to 60th levels, and the 20 levels corresponding to the picture object glare are the 61st to the 61st. At the 80th level, the 20 levels corresponding to the contrast of the picture object are the 81st to the 100th level, and the level corresponding to the picture object is the 101st to 102th level, and the ratio of the main area of the picture object corresponds to the 20th level of the 103th to the 122th level. The level corresponding to the color scheme of the main body of the picture object is the 123th to the 125th level, and the level corresponding to the color depth of the main body of the picture object is the 126th to the 127th level, and each level corresponds to the corresponding accurate visual feature value of the specified numerical range. Then, for the case where the visual feature extracted by the extracting unit 101 is a hierarchical visual feature value, the extracting unit 101 as shown in FIG. 11 may include:

An extracting subunit 1011, configured to extract an accurate visual feature value of a visual feature of the acquired picture object;

The hierarchical sub-unit 1012 is configured to determine a hierarchical visual feature value to which the precise visual feature value belongs as a visual feature of the acquired picture object.

Specifically, the number determining unit 102 may determine the number of picture objects browsed by the user having the same visual feature, and the number of picture objects selected by the user having the same visual feature. It can be seen that the number of picture objects having the same visual feature in the acquired picture object is determined, that is, the number of picture objects corresponding to the same visual feature involved in the user behavior record is determined.

An embodiment of the user object record that is acquired by the acquisition unit 100 includes only the scene of the picture object selected by the user. The structure of the preference score determining unit 103 is as shown in FIG. 12, and includes:

a first base score determining sub-unit 1031 for determining a base score of the visual feature based on a number of picture objects having the same visual feature and a correction value, wherein the correction value indicates that the picture object having the visual feature is in the picture object The ratio in the library. Taking the color depth of the main body of the picture object as an example, the correction value includes the ratio of the picture object whose color is dark in the picture object library in the picture object library, and the picture object whose picture object color is light color. The ratio of the image object library. The library of picture objects may be, for example, a database of all picture objects that are searchable by a server on a particular website. As shown in Figure 3, the picture object master is determined. Schematic diagram of the method of correcting the body color shade. Assume that the picture object library contains four picture objects, the ID range is 1000 to 1003, and the picture object body color of the picture object whose ID is 1000, 1001, 1003 is dark, and the picture object body of the picture object with the ID of 1002 The color is light, then the correction value of the color of the main body of the picture object in the picture object library is 0.75 (3/4, that is, the number of picture objects whose color is dark in the picture object body/the number of picture objects in the picture object library) The correction value of the light color of the main body of the picture object is 0.25 (1/4, and the number of picture objects whose picture object color is light color/the number of picture objects in the picture object library). The same method can be used to determine the correction values of other visual features. It can be understood that the correction value can be calculated once at a fixed time interval.

The first preference score determining sub-unit 1032 is configured to determine a user's preference score for the visual feature based on the base score of the visual feature. After the first base score determining sub-unit 1031 determines the base score of each visual feature, the first preference score determining sub-unit 1032 may determine the user's preference score for the visual feature by using various existing weight calculating methods, for example, The LR (Logistic Regression) model is used for training, and the embodiment of the present application does not specifically limit this.

Another embodiment is that the picture object related to the user behavior record acquired by the obtaining unit 100 includes a picture object browsed by the user and a picture object selected by the user, and the structure of the preference score determining unit 103 is as shown in FIG. 13 and includes :

a second base score determining sub-unit 1033, configured to determine a base score of the visual feature based on a number of picture objects having the same visual feature and a corresponding weighting factor; wherein the weighting factor is determined according to whether the behavior record is browsing or selecting;

The second preference score determination sub-unit 1034 is configured to determine a user's preference score for the visual feature based on the base score of the visual feature.

In this embodiment, the picture object involved in the user behavior record includes the picture object browsed by the user and the picture object selected by the user, and the user's interest level with the selected picture object is higher than the interest level of the browsed picture object, so the second basic score The determining sub-unit 1033 may determine different weighting factors for the picture object browsed by the user and the picture object selected by the user. When determining the score of the visual feature, distinguishing the picture object browsed by the user from the picture object selected by the user, respectively multiplied by the corresponding weighting factor.

By differently treating the picture object selected by the user and the picture object browsed by the user to determine the basic score of the visual feature, the user's interest point can be more prepared.

The second preference score determination sub-unit 1034 can determine the user's preference score for the visual feature by using various weight calculation methods. For example, the LR (Logistic Regression) model can be used for training. There are no specific restrictions on this.

The embodiment of the present application further provides an apparatus for sorting a search result of a picture object corresponding to the above method for sorting the search result of the picture object. The structure of the device is as shown in FIG. 14 , and the device mainly includes:

a picture object obtaining unit 140, configured to acquire the search request in response to receiving a search request of the user A matching picture object; the method of the picture object obtaining unit 140 acquiring the picture object matching the search request is not specifically limited. The search result matching the search request may be first obtained, and then the picture object corresponding to the search result is obtained.

The visual feature extraction unit 141 is configured to extract a visual feature of the acquired picture object; the visual feature includes at least one of the following:

Picture object brightness, picture object saturation, picture object sharpness, picture object contrast, picture object brilliance, picture object puzzle, picture object body area ratio, picture object body color depth, picture object body color scheme. The method for extracting the visual features of the visual features described above by the visual feature extraction unit 141 is the same as that described in the above embodiments, and details are not described herein again.

It should be noted that, in this embodiment, the visual feature extraction unit 141 extracts whether the visual feature of the acquired picture object is an accurate visual feature value or a hierarchical visual feature value, which is consistent with the visual feature acquired when determining the user's preference score for the visual feature. That is, if the visual feature acquired when determining the user's preference score for the visual feature is the accurate visual feature value, then this step also needs to extract the accurate visual feature value of the acquired visual feature of the picture object, and vice versa. Hierarchical visual feature values.

The comprehensive preference score obtaining unit 142 is configured to obtain a comprehensive preference score of the user for the acquired picture object based on the user's preference score for the visual feature;

The user's preference score for the visual feature may be determined by the above device for sorting the search results of the auxiliary picture object, and details are not described herein;

In one embodiment, the integrated preference score obtaining unit 142 uses the sum of the user's preference scores for the visual features as a comprehensive preference score of the user for the acquired picture object.

In another embodiment, the picture object obtaining unit 140 is further configured to acquire an original sort score of the picture object that matches the search request. Since the method of the embodiment of the present application is not used, when the website searches for a picture object that matches the search request, when the website server gives the matched picture object, the picture object is given to the picture object according to the text description of the picture object, as in the prior art. An inherent sorting, which is also often based on the sorting score of each picture object, except that the sorting score is obtained based on the degree of matching between the text description of the picture object and the keywords in the search request. The ranking score is referred to as an original ranking score, that is, the original ranking score is a ranking score assigned to the search result without considering the user's preference score for the visual feature of the embodiment. In this embodiment, the original ranking score of the picture object matching the search request is acquired while acquiring the search result matching the request; after that, the integrated preference score obtaining unit 142 compares each visual feature of the picture object according to the user pair. And selecting a weighted sum of the preference weights of the visual features; and adding the weighted sum to the original of the acquired image objects The scores are sorted, and the obtained sum values are used as the user's interest scores for the acquired image objects. In this embodiment, when the search result of the picture object is sorted, the preference weight preference score of the user for the visual feature and the matching degree of the search result with the search request are considered, and the search result obtained by sorting the search result according to the interest score comprehensive preference score is more capable. Improve the user experience.

Since the preference score value obtained according to the embodiment is a value less than 1, the user's preference score for the visual feature plus the original value ranking score has less influence on the original value ranking score, in order to reflect the user's vision. The effect of the preference score of the feature on the original ranking score, another embodiment of the present application, the integrated preference score obtaining unit 142 may multiply the sum of the user's preference scores of the visual features by a predetermined adjustment value, and obtain the product plus the obtained The sum value obtained after the original sort score is used as the composite preference score. The predetermined adjustment value may be, for example, 100, or 10, or the like. Assuming that the original ranking score is 10 and the user's preference score for the visual feature is 0.06, then the sum of the original ranking score and the user's preference score for the visual feature is 10+0.06=10.06, if the user is visually The sum of the feature's preference scores is multiplied by a predetermined adjusted value of 100, and then added to the original sorted score is 10+0.06*100=16. It can be seen that the integrated preference score calculated by the latter has more influence on the original sorted score. Big.

In this embodiment, when the search result of the picture object is sorted, the user's preference score for the visual feature and the matching degree between the search result and the search request are considered, and then the search result obtained by sorting the search result according to the integrated preference score can further improve the user experience.

The sorting unit 143 is configured to sort the picture objects according to a comprehensive preference score of the acquired picture object by the user. The picture objects may be sorted according to the user's comprehensive preference scores of the acquired picture objects from high to low.

The embodiment may determine the user's comprehensive preference score for the acquired picture object based on the visual feature of the search result and the user's preference score for the visual feature, and according to the user's comprehensive preference score for the acquired picture object. The picture object sorting, because the user's preference score for the visual feature determined based on the visual feature of the picture object can more accurately express the user's interest point, the sorting of the picture object based on the image object can be more convenient for the user to find the required search result. , thereby reducing the traffic consumption caused by the user repeatedly selecting the required search results.

It should be noted that the present invention can be implemented in software and/or a combination of software and hardware, for example, using an application specific integrated circuit (ASIC), a general purpose computer, or any other similar hardware device. In one embodiment, the software program of the present invention may be executed by a processor to implement the steps or functions described above. Likewise, the software program (including related data structures) of the present invention can be stored in a computer readable recording medium such as a RAM memory, a magnetic or optical drive or a floppy disk and the like. Additionally, some of the steps or functions of the present invention may be implemented in hardware, for example, as a circuit that cooperates with a processor to perform various steps or functions.

Additionally, a portion of the invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide a method and/or solution in accordance with the present invention. The program instructions for invoking the method of the present invention may be stored in a fixed or removable recording medium and/or transmitted by a data stream in a broadcast or other signal bearing medium, and/or stored in a The working memory of the computer device in which the program instructions are run. Herein, an embodiment in accordance with the present invention includes a device including a memory for storing computer program instructions and a processor for executing program instructions, wherein when the computer program instructions are executed by the processor, triggering The apparatus operates based on the aforementioned methods and/or technical solutions in accordance with various embodiments of the present invention.

It is apparent to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, and the present invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the invention is defined by the appended claims instead All changes in the meaning and scope of equivalent elements are included in the present invention. Any reference signs in the claims should not be construed as limiting the claim. In addition, it is to be understood that the word "comprising" does not exclude other elements or steps. A plurality of units or devices recited in the system claims can also be implemented by a unit or device by software or hardware. The first, second, etc. words are used to denote names and do not denote any particular order.

Claims

A method for assisting in sorting search results of picture objects, comprising:

Obtaining the image object involved in the user behavior record within the set time range;

Extracting visual features of the acquired picture object;

Determining the number of picture objects having the same visual feature in the acquired picture object;

Based on the number of picture objects having the same visual feature, the user's preference score for the visual feature is determined for ordering the picture object search results.
The method of claim 1 wherein said visual features comprise at least one of the following:

Picture object brightness, picture object saturation, picture object sharpness, picture object contrast, picture object brilliance, picture object puzzle, picture object body area ratio, picture object body color depth, picture object body color scheme.
The method of claim 1 wherein determining a user's preference score for the visual feature based on the number of picture objects having the same visual feature comprises:

Determining a base score of the visual feature based on a number of picture objects having the same visual feature and a correction value, wherein the correction value represents a ratio of the picture object having the visual feature in the picture object library;

A user's preference score for the visual feature is determined based on the base score of the visual feature.
The method of claim 1 or 3 wherein said behavior record comprises a selection.
The method of claim 1, wherein the behavior record comprises: browsing and selecting, determining a user's preference score for the visual feature based on the number of picture objects having the same visual feature comprises:

Determining a base score of the visual feature based on a number of picture objects having the same visual feature and a corresponding weighting factor, wherein the weighting factor is determined based on whether the behavior record is browsed or selected;

A user's preference score for the visual feature is determined based on the base score of the visual feature.
The method of claim 1 wherein said visual features are characterized by precise visual feature values.
The method of claim 1 wherein the visual feature is characterized by an accurate visual feature value or a hierarchical visual feature value, wherein wherein the acquired visual image feature is extracted by the hierarchical visual feature value criterion Visual features include:

Extracting precise visual feature values of the visual features of the acquired picture object;

A hierarchical visual feature value to which the precise visual feature value belongs is determined as a visual feature of the acquired picture object.
A method for sorting search results of picture objects, comprising:

Acquiring a picture object that matches the search request in response to receiving a search request of the user;

Extracting visual features of the acquired picture object;

Obtaining a comprehensive preference score of the user for the acquired picture object based on a user's preference score for the visual feature, wherein the user's preference score for the visual feature is for assisting according to any one of claims 1 to 7 The method of sorting the search results of the picture object is determined;

The picture objects are sorted according to a user's comprehensive preference score for the acquired picture object.
The method of claim 8 wherein said visual features comprise at least one of the following:

Picture object brightness, picture object saturation, picture object sharpness, picture object contrast, picture object brilliance, picture object puzzle, picture object body area ratio, picture object body color depth, picture object body color scheme.
The method of claim 8 wherein obtaining a user's integrated preference score for the acquired picture object based on a user's preference score for the visual feature comprises:

The sum of the user's preference scores for the visual features is taken as the user's overall preference score for the acquired picture object.
The method of claim 8, wherein acquiring the picture object that matches the search request further comprises: acquiring an original ranking score of the picture object that matches the search request, based on the user's The preference scores obtained by the user for the integrated preference scores of the acquired image objects include:

Adding the sum of the preference scores of the user to the visual feature to the obtained original sort score of the image object as a comprehensive preference score of the user for the acquired image object; or

The sum value obtained by adding the sum of the preference scores of the visual features to the predetermined adjustment value plus the obtained original sort score of the picture object is used as a comprehensive preference score of the user for the acquired picture object.
The method according to any one of claims 8, 10, 11, wherein sorting the picture objects according to a user's comprehensive preference score for the acquired picture object comprises:

The picture objects are sorted in descending order of the user's overall preference scores for the acquired picture objects.
An apparatus for assisting in sorting search results of picture objects, comprising:

An obtaining unit, configured to acquire a picture object involved in a user behavior record in a set time range;

An extracting unit, configured to extract a visual feature of the acquired picture object;

a number determining unit, configured to determine a number of picture objects having the same visual feature among the acquired picture objects;

And a preference score determining unit, configured to determine a user's preference score for the visual feature based on the number of picture objects having the same visual feature, for ordering the picture object search result.
The device of claim 13 wherein said visual feature comprises at least one of the following:

Picture object brightness, picture object saturation, picture object sharpness, picture object contrast, picture object brilliance, picture object puzzle, picture object body area ratio, picture object body color depth, picture object body color scheme.
The apparatus according to claim 13, wherein the preference score determining unit comprises:

a first base score determining subunit for determining a base score of the visual feature based on a number of picture objects having the same visual feature and a correction value, wherein the correction value indicates a ratio of the picture object having the visual feature in the picture object library ;

The first preference score determining subunit is configured to determine a user's preference score for the visual feature based on the base score of the visual feature.
Apparatus according to claim 13 or claim 15 wherein said behavior record comprises a selection.
The apparatus according to claim 13, wherein said behavior record comprises: browsing and selecting, and the preference score determining unit comprises:

a second base score determining subunit, configured to determine a base score of the visual feature based on a number of picture objects having the same visual feature and a corresponding weighting factor, wherein the weighting factor is determined according to whether the behavior record is browsed or selected;

A second preference score determining subunit is configured to determine a user's preference score for the visual feature based on the base score of the visual feature.
The apparatus of claim 13 wherein said visual features are characterized by precise visual feature values.
The apparatus of claim 13 wherein said visual features are characterized by precise visual feature values or hierarchical visual feature values, wherein in the case where the visual features are characterized by hierarchical visual feature values, the extracting unit comprises:

Extracting a subunit for extracting an accurate visual feature value of a visual feature of the acquired picture object;

The hierarchical sub-unit is configured to determine a hierarchical visual feature value to which the precise visual feature value belongs as a visual feature of the acquired picture object.
An apparatus for sorting search results of picture objects, comprising:

a picture object obtaining unit, configured to acquire a picture object that matches the search request in response to receiving a search request of the user;

a visual feature extraction unit, configured to extract a visual feature of the acquired picture object;

a comprehensive preference score obtaining unit, configured to obtain a user's comprehensive preference score for the acquired picture object based on a user's preference score for the visual feature, wherein the user's preference score for the visual feature is determined by claims 13 to 19 Determining, by a device for assisting the ranking of search results of a picture object;

a sorting unit, configured to sort the picture objects according to a user's comprehensive preference score of the acquired picture objects.
The device of claim 20 wherein said visual feature comprises at least one of the following:

Picture object brightness, picture object saturation, picture object sharpness, picture object contrast, picture object brilliance, picture object puzzle, picture object body area ratio, picture object body color depth, picture object body color scheme.
The apparatus according to claim 20, wherein the integrated preference score obtaining unit is configured to:

The sum of the user's preference scores for the visual features is taken as the user's overall preference score for the acquired picture object.
The device according to claim 20, wherein the picture object obtaining unit is further configured to: obtain an original sort score of the picture object that matches the search request, and the integrated preference score obtaining unit is configured to:

Adding the sum of the preference scores of the user to the visual feature to the obtained original sort score of the image object as a comprehensive preference score of the user for the acquired image object; or

The sum value obtained by adding the sum of the preference scores of the visual features to the predetermined adjustment value plus the obtained original sort score of the picture object is used as a comprehensive preference score of the user for the acquired picture object.
The apparatus according to any one of claims 20, 22, 23, wherein the sorting unit is configured to:

The picture objects are sorted in descending order of the user's overall preference scores for the acquired picture objects.