An image trading platform management system based on artificial intelligence

Abstract

The present invention belongs to the technical field of picture transaction management, and specifically relates to a picture transaction platform management system based on artificial intelligence. By setting an upload entry in the picture display page of the transaction platform, historical purchasers are allowed to upload the usage status image and usage evaluation of the picture on the picture details page, thereby using the historical usage evaluation of the picture to screen the picture when performing picture retrieval, and can help purchasers better evaluate whether the picture is suitable for their needs by displaying the real usage status image and evaluation, thereby improving the accuracy and satisfaction of the selection. In addition, by increasing the upload of the usage status image and usage evaluation of the historical transaction pictures, the usage scenario is extracted from the usage status image with effective and positive evaluation and updated in the picture details page, thereby realizing the dynamic update of the picture application scenario, and can discover the potential value and new application scenario of the picture in actual application through the analysis of the actual usage scenario.

Description

Technical Field

[0001] The present invention belongs to the technical field of picture transaction management, and in particular relates to a picture transaction platform management system based on artificial intelligence. Background Art

[0002] With the advent of the digital age, the convenience of content production and dissemination has been significantly improved. Digital platforms such as websites, social media, and mobile applications have seen a surge in demand for high-quality image materials to attract user attention and improve user experience. Against this backdrop, image trading platforms have emerged, providing efficient one-stop online services that enable buyers to quickly locate and obtain the image resources they need without being restricted by time and space. At the same time, these platforms have also simplified the copyright licensing and payment processes, greatly improving the efficiency and convenience of transactions.

[0003] In the prior art, a Chinese invention patent with a publication number of CN101556584A discloses a computer system and method for realizing image transactions, which classifies images uploaded from the Internet or other means into the image database and refreshes the image classification device, and then the user logs in to the system to retrieve the images stored in the image database through a retrieval device using multiple image characteristics as retrieval conditions, and selects the required images for ordering. The invention can quickly select the required images by classifying the uploaded images and setting multiple retrieval conditions. However, the images stored in the image database of the invention are limited to the images themselves, lacking the actual display of application scenarios. However, the main purpose of image purchase is to use them in specific scenarios. Simply relying on the original images displayed in the database is difficult to help buyers accurately evaluate whether the performance of the images in actual applications meets the expected needs, resulting in buyers having difficulty making accurate choices without reference to actual application scenarios, which invisibly increases the difficulty of decision-making and reduces the purchase conversion rate to a certain extent.

[0004] In addition, the invention usually classifies images in the image database based on their known attributes, which is a static classification method. Due to the lack of in-depth analysis and research on historical transaction data, this classification method is difficult to capture new expanded uses of images in practical applications in a timely manner. For example, the classification of application scenarios is often based on the original creative intent of the image creator. However, in actual applications, it is often the case that buyers use images for scenarios other than the original intent and the results are equally good. This shows that this classification method may be too rigid and cannot fully reflect the potential value of images in diversified application scenarios. When users are looking for images, they may find it difficult to find images that meet their specific needs because the classification is too rigid, which reduces user experience and satisfaction. Summary of the invention

[0005] In view of this, the present invention aims to propose an image trading platform management system based on artificial intelligence, which can effectively solve the deficiencies in the prior art by optimizing the classification and retrieval of images on the image trading platform according to application scenarios.

[0006] The purpose of the present invention can be achieved through the following technical solutions: an artificial intelligence-based picture trading platform management system, comprising the following modules: a picture display upload module, used to set an upload entry in the picture display page of the trading platform, allowing historical purchasers to upload picture usage status images and usage evaluations on the picture details page.

[0007] The usage scenario dynamic update module is used to regularly identify valid positive comments from the usage status images and usage evaluations uploaded to the picture details page according to the set update cycle, and extract usage scenarios from the usage status images of valid positive comments, and then update the extracted usage scenarios in the picture details page.

[0008] The image retrieval module is used to provide usage scenario retrieval conditions on the image retrieval page of the trading platform, allowing users to enter usage scenarios for image retrieval.

[0009] The first picture screening module is used to preliminarily screen out pictures that meet user needs from each picture detail page in the picture display page according to the usage scenario input by the user, record them as candidate pictures, and retrieve the usage evaluation of the candidate pictures, thereby deeply screening out the available pictures.

[0010] The second picture screening module is used to identify the attention features of the effective positive comments on the available pictures, and then retrieve the user's historical transaction records to parse the user's tendency to pay attention to the features, thereby matching the attention features of the available picture usage evaluations with the user's tendency to pay attention to the features to screen out suitable pictures.

[0011] The picture recommendation display module is used to interactively display the selected suitable pictures.

[0012] The scene configuration library is used to store background objects that are commonly configured for different scenes.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention sets an upload entry in the picture display page of the trading platform to allow historical purchasers to upload the usage status image and usage evaluation of the picture on the picture details page, thereby using the historical usage evaluation of the picture to screen the pictures when performing picture retrieval, which can help purchasers better evaluate whether the picture is suitable for their needs by displaying the real usage status image and evaluation, thereby improving the accuracy and satisfaction of the selection. At the same time, the usage evaluation and status image of the historical purchasers provide an additional source of information, making the picture information richer and more comprehensive.

[0014] 2. The present invention increases the upload of usage status images and usage evaluations of historical transaction images, and then extracts usage scenarios from the usage status images with valid and positive evaluations and updates them in the picture details page, thereby realizing dynamic update of picture application scenarios. By analyzing the actual usage scenarios, the present invention can discover the potential value and new application scenarios of pictures in actual applications. On the one hand, it provides buyers with a broader and more referenceable range of picture screening, and on the other hand, it helps creators and platforms to better understand and promote pictures.

[0015] 3. The present invention adopts an intelligent recommendation method when screening pictures from the usage status images and usage evaluations of historical transaction pictures uploaded on the picture display page based on the application scenarios input by the purchaser. The purchaser does not need to view and screen them one by one, which greatly saves time and energy, makes the purchaser's experience smoother and more pleasant, and enhances the user's favorability towards the platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0017] Figure 1 It is a schematic diagram of the connection of each module of the system of the present invention.

[0018] Figure 2 Implementation flow chart of dynamic update of scene usage in the present invention

[0019] Figure 3 This is a flowchart for implementing the first screening of images in the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] The present invention proposes an artificial intelligence-based picture transaction platform management system, which includes a picture display upload module, a usage scenario dynamic update module, a picture retrieval module, a picture first screening module, a picture second screening module, a picture recommendation display module and a scene configuration library.

[0022] See also Figure 1As shown, the above-mentioned picture display upload module is connected to the usage scenario dynamic update module, the usage scenario dynamic update module and the picture retrieval module are both connected to the picture first screening module, the picture first screening module is connected to the picture second screening module, the picture second screening module is connected to the picture recommendation display module, and the scene configuration library is connected to the usage scenario dynamic update module.

[0023] The picture display upload module is used to set an upload entry in the picture display page of the trading platform, allowing historical buyers to upload the picture's usage status image and usage evaluation on the picture details page.

[0024] In the specific implementation of the above solution, a functional module is integrated into the picture details page of the trading platform, such as a "share use case" button or link, to guide users to upload their use status images and use evaluations. The use evaluations include the evaluation content of historical buyers and the evaluation star rating generated by the system for the evaluation content.

[0025] The usage scenario dynamic update module is used to regularly identify valid positive comments from the usage status images and usage comments uploaded by users on the picture details page according to the set update cycle, and extract usage scenarios from the usage status images of valid positive comments, and then update the extracted usage scenarios in the picture details page. Figure 2 shown.

[0026] Applied to the above scheme, the update cycle setting can be dynamically adjusted, and is specifically set according to the frequency of generating transaction records in the trading platform, that is, the transaction frequency. When the transaction frequency is high, it means that the usage status images and evaluations uploaded by the purchaser are updated faster. At this time, a shorter update cycle can be set to reflect the latest user feedback more quickly. When the transaction frequency is low, it means that the usage status images and evaluations uploaded by the user are updated slowly. A longer update cycle can be set to avoid unnecessary resource consumption. In this way, the usage evaluation retrieved from the picture details page can always be the latest evaluation while avoiding excessive resource consumption, thereby reducing the proportion of outdated information and making the usage evaluation retrieved from the picture details page more accurate.

[0027] In the specific implementation of the above solution, the following identification process is effectively and positively evaluated: the usage evaluation uploaded by each picture in the picture display page within the update period is mapped with the usage status image to obtain the usage status image of each usage evaluation mapping.

[0028] As a specific example of the above implementation, the mapping of the usage evaluation and the usage status image can associate the usage evaluation with the usage status image through an identifier (such as purchaser ID+image ID+evaluation ID).

[0029] Each usage evaluation is segmented into several participles, and each segmented participle is tagged with a part of speech.

[0030] The above-mentioned word segmentation can be performed using a Chinese word segmentation tool, and the part-of-speech tagging can be performed using a part-of-speech tagging tool, where the parts of speech include but are not limited to nouns, adjectives, adverbs, and the like.

[0031] Evaluation representation phrases are extracted based on the part of speech of each participle. If an evaluation representation phrase is not extracted from a usage evaluation, the usage evaluation is removed and the remaining usage evaluation is recorded as a valid evaluation.

[0032] In the preferred implementation of the above scheme, the evaluation representation phrases are refined as follows: adjectives and adverbs are collectively referred to as description words, and the participles with noun and description words are selected according to the part of speech of each participle and recorded as noun participles and description participles respectively.

[0033] It is important to know that adjectives and adverbs are called descriptive words because they are mainly used in language to describe and modify nouns, verbs or other words, providing detailed information or characteristics about objects or actions. In evaluation, adjectives and adverbs can provide detailed information about the evaluation object and become an important part of expressing the evaluation. By using these words, the evaluator can convey his or her views and feelings more accurately, making the evaluation more expressive and persuasive.

[0034] Adjacent noun participles and descriptive word participles are combined into phrases as evaluation representation phrases.

[0035] In the example implemented above, suppose the evaluation is "This picture is used for office wall decoration. The colors are relatively bright and the contrast is very strong." The word segmentations divided by the Chinese word segmentation tool are "this", "picture", "used for", "office", "wall", "decoration", "color", "relatively", "bright", "contrast", "very", and "strong". Among them, the descriptive words are "relatively", "bright", "very", and "strong", and the nouns are "office", "wall", "color", and "contrast". The adjacent noun participles and descriptive word participles are "color", "relatively", "bright" and "contrast", "very", and "strong". The evaluation representation phrases in this example are "the colors are relatively bright" and "the contrast is very strong".

[0036] The present invention identifies effective evaluations based on whether evaluation representation phrases can be extracted from usage evaluations, and eliminates usage evaluations that do not have evaluation representation phrases, thereby ensuring that the remaining evaluations contain substantive information, thereby improving the accuracy and reliability of subsequent analysis, and helping to ensure that the analysis conclusions are based on valuable user feedback. At the same time, eliminating invalid evaluations can reduce the computing resources and storage space required for processing and storing useless data.

[0037] The rating stars are extracted from the valid reviews, and the emotional tendency corresponding to the valid reviews is obtained based on the corresponding relationship between the rating stars and the emotional tendency. In the specific implementation, a mapping table of rating stars and emotional tendency can be established. As shown in Table 1, when the rating star is 3 stars or above, the emotional tendency corresponding to the evaluation is positive.

[0038] Table 1

[0039]

[0040] It is important to understand that 3-star reviews are often considered the boundary of neutral preference on many e-commerce platforms, app stores, etc. Reviews below 3 stars are often considered negative feedback, while 3 stars and above are considered positive feedback.

[0041] From the sentiment tendencies corresponding to the valid evaluations, the valid evaluations with positive sentiment tendencies are selected as the valid positive evaluations.

[0042] By identifying effective positive evaluations of the usage evaluations uploaded by historical purchasers, the present invention can more accurately understand which images perform well in actual applications, thereby providing more accurate recommendations for new users.

[0043] In a further preferred implementation, the use scene is extracted from the use state image with effective positive evaluation in the following process: background separation is performed on the use state image with effective positive evaluation, and background objects are identified from the separated background.

[0044] The purpose of the background segmentation mentioned above is to separate the foreground (i.e., the picture used by the user) from the background in the usage state image, so as to better identify the background objects and understand the specific scene the image is in. For example, the background separation can be performed by edge detection, background subtraction, etc.

[0045] The number of recognized background objects is counted, and each background object is matched with the background objects conventionally equipped for different scenes in the scene configuration library, thereby obtaining a scene in which each background object matches.

[0046] Applied to the above embodiment, different scenes in the scene configuration library are usually equipped with specific background objects to help users better understand and apply. For example, objects usually equipped in interior design scenes include sofas, televisions, coffee tables, etc., objects usually equipped in office design scenes include desks, computers, filing cabinets, etc., and objects usually equipped in commercial display scenes include display stands, shelves, etc.

[0047] The number of recognized background objects is counted, and each background object is matched with the background objects conventionally equipped for different scenes in the scene configuration library, thereby obtaining a scene in which each background object matches.

[0048] The scenes matching the background objects are classified as the same scenes, the occurrence frequency of each scene is counted, and then the scene corresponding to the maximum occurrence frequency is used as the usage scene for effectively and positively evaluating the corresponding usage status image.

[0049] In the example of the above solution, if objects such as a sofa, a television, a coffee table, and a computer are recognized, it may be an interior design scene.

[0050] The image retrieval module is used to provide usage scenario retrieval conditions on the image retrieval page of the trading platform, allowing the user to enter the usage scenario for image retrieval.

[0051] The first image screening module is used to preliminarily screen out images that meet the user's needs from each image detail page in the image display page according to the usage scenario input by the user, and record them as candidate images, and further retrieve the usage evaluation of the candidate images, thereby deeply screening out the available images. Figure 3 shown.

[0052] In a possible implementation of the above scheme, the screening process of candidate images is to compare the usage scenario input by the user with the usage scenarios updated in each image detail page, and extract the images that meet the usage scenario input by the user as candidate images.

[0053] It should be understood that the alternative images are selected based on the usage scenarios input by the user, which can ensure that the selected images are selected based on the actual needs of the user.

[0054] In a further possible implementation, the process of screening available images is as follows: the proportion of valid positive reviews in the usage reviews uploaded by historical purchasers corresponding to the alternative images is counted, and compared with the set standard ratio. For example, the standard ratio is 80%, and the alternative images that meet the standard ratio are extracted as reference images.

[0055] It should be understood that after screening out candidate images, reference images are further screened out based on the proportion of valid positive reviews, so that the available images are obtained based on valid positive reviews, which can improve the reliability of recommendation results and user satisfaction.

[0056] The rating stars are extracted from the usage reviews of the reference images, and the rating stars of each usage review corresponding to each reference image are compared to obtain the median rating star and the average rating star of each reference image.

[0057] It should be added that the average rating reflects the overall level, while the median rating better reflects the stability of the middle value.

[0058] The median average deviation of each reference image is obtained by taking the absolute value of the difference between the median rating and the average rating of each reference image and dividing it by the median rating.

[0059] The median average deviation of each reference image is compared with the system-limited deviation. For example, the limited deviation is 0.3. If the median average deviation of a reference image reaches the system-limited deviation, it indicates that there is a large gap between the median rating star and the average rating star, reflecting that the rating star distribution of the evaluations used in the reference image is relatively scattered, and there is a risk of polarization of evaluations. At this time, the reference image is eliminated. Conversely, if the median average deviation of a reference image does not reach the system-limited deviation, it indicates that there is a small gap between the median rating star and the average rating star, reflecting that the rating star distribution of the evaluations used in the reference image is relatively even, which means that the evaluations of historical buyers tend to be consistent, which means that historical buyers have reached a consensus on the evaluations of the images.

[0060] Through the above operation, the present invention makes the available pictures obtained by screening to be pictures with a relatively uniform distribution of star ratings, which can improve the quality of recommended pictures and ensure that the recommended pictures have good evaluations among most users, making the pictures more stable when recommended and not affecting the overall recommendation results due to fluctuations in individual evaluations. In other words, eliminating pictures with dispersed distribution of star ratings can reduce the risk of extreme evaluations in the recommendation results and reduce the confusion of users caused by inconsistent evaluations.

[0061] The second picture screening module is used to identify the attention features of the effective positive evaluations of the available pictures, and then retrieve the user's historical transaction records to parse the user's tendency to pay attention to the features, thereby matching the attention features of the available picture usage evaluations with the user's tendency to pay attention to the features, and screening out suitable pictures.

[0062] In a preferred implementation, the following process is performed to identify the features of the effective positive comments on the available pictures: extract the effective positive comments from the usage comments of the available pictures, and separate the comment feature and the comment description from the comment representation phrases extracted from the comments.

[0063] It should be noted that the evaluation feature is the evaluation object. In the example of the present invention using the evaluation "This picture is used for office wall decoration, the color is bright and the contrast is very strong", the evaluation feature of the evaluation representation phrase "the color is bright" is split into color, and the evaluation description is bright, and the evaluation feature of "the contrast is very strong" is split into contrast, and the evaluation description is very strong.

[0064] The extracted evaluation representation phrases are numbered in the order in which they appear in the usage evaluation, and the numbers are used as the occurrence numbers of the evaluation features obtained by splitting the evaluation representation phrases.

[0065] In the above example, “the color is brighter” appears first and “the contrast is stronger” appears later, so the appearance numbers of the evaluation features “color” and “contrast” are 1 and 2.

[0066] It should be pointed out that, since there is a one-to-one correspondence between the evaluation representation phrase and the evaluation feature, the appearance number of the evaluation representation phrase in the evaluation is the appearance number of the evaluation feature.

[0067] Degree words are extracted from the evaluation descriptions split from each evaluation representation phrase, and the weight values ​​corresponding to the degree words are obtained according to the pre-defined weight correspondence between the degree words and the sentiment intensity.

[0068] It should be noted that in evaluation descriptions, degree words usually refer to words used to express the degree or intensity of a certain characteristic, state or behavior, for example, "very", "relatively", "very", "slightly", etc. In the evaluation representation phrases in the above examples, "the color is brighter" and "the contrast is very strong" contain degree words "relatively" and "very" respectively.

[0069] What needs to be further known is that the higher the sentiment intensity of the degree word, the greater the corresponding weight value. Specifically, the corresponding weight relationship between the degree word and the sentiment intensity can be set according to the sentiment intensity size relationship of the degree words commonly used in the Chinese language. For example, the sentiment intensity size relationship corresponding to "very", "relatively", "very", and "slightly" is slightly < relatively < very < very. Under such a relationship, the weight range is set between 0 and 1. At this time, the corresponding weights of "very", "relatively", "very", and "slightly" are 0.8, 0.4, 0.6, and 0.2 respectively.

[0070] The weight values ​​corresponding to each degree word in the evaluation description corresponding to each evaluation representation phrase are accumulated to obtain the sentiment intensity coefficient of the evaluation feature corresponding to the evaluation representation phrase.

[0071] In the above example, since "the color is brighter" and "the contrast is very strong" both contain only one degree word, the sentiment intensity coefficients of the evaluation features "color" and "contrast" corresponding to "the color is brighter" and "the contrast is very strong" are 0.4 and 0.6 respectively.

[0072] Substitute the occurrence number and sentiment intensity coefficient corresponding to each evaluation feature in the available pictures into the formula

[0073] Calculate the attention corresponding to each evaluation feature in the available image's effective positive evaluation , where Indicates the occurrence number of the evaluation feature, represents the number of evaluation characterization phrases, where the number of evaluation characterization phrases is the number of evaluation features. Represents the sentiment intensity coefficient of the evaluation feature.

[0074] From the above calculation formula of attention, we can know that the earlier the evaluation feature appears and the larger the emotional intensity coefficient is, the greater the attention is. This is because when users write reviews, they usually first mention the features they care about most or are most impressed by. Therefore, the appearance of evaluation features in the front often means that users think these features are more important. The emotional intensity coefficient reflects the user's emotional intensity for the feature. The greater the intensity, the stronger the user's feelings about the feature.

[0075] The present invention quantifies attention by using the order of appearance and sentiment intensity of each evaluation feature in the effective positive evaluation, so as to analyze the evaluation content of the user in more detail and understand the specific features that the user pays attention to.

[0076] Classify the evaluation features existing in all valid positive evaluations of available images according to the same evaluation features, and obtain a number of valid positive evaluations corresponding to each evaluation feature;

[0077] The attention degree of each evaluation feature in different valid positive evaluations is accumulated and calculated to obtain the total attention degree of each evaluation feature, and then the evaluation feature corresponding to the maximum total attention degree is extracted as the attention feature of the available image.

[0078] In a further preferred implementation, the user tends to pay attention to the following analysis process: when the user has a historical transaction record, the transaction image is extracted from the user's historical transaction record, and the usage evaluation of the transaction image is retrieved from the detail page of the transaction image.

[0079] Similarly, effective positive comments are screened out from the usage evaluations of transaction images, and attention features are identified for the effective positive comments to obtain attention features of transaction images.

[0080] The attention features of each historical transaction record and the corresponding transaction image are compared, and the attention features with the highest frequency of occurrence are extracted as the user's preferred attention features.

[0081] In further innovative implementations, the user tendency attention feature also includes the following parsing process: when the user has no historical transaction records, the historical transaction records are retrieved based on the usage scenarios input by the user, and then the historical transaction records that meet the usage scenarios input by the user are extracted from the retrieval records involved in the historical transaction records and recorded as associated historical transaction records.

[0082] The associated historical transaction records are similarly analyzed according to the analysis process of the features that the user tends to pay attention to when the user has historical transaction records.

[0083] In a further preferred implementation, the adapted pictures are screened as follows: the attention features of the available pictures are matched with the attention features that the user tends to pay attention to, and the available pictures that are successfully matched are screened out as the adapted pictures.

[0084] The present invention analyzes the user's tendency to pay attention to features as long as it takes into account that the user is more willing to receive recommendations that match their interests. Therefore, by identifying the user's tendency to pay attention to features, more personalized recommendations are provided to the user, which is conducive to improving user satisfaction.

[0085] It should be pointed out that when performing adaptive image screening based on the usage scenarios input by the user, a group of preliminary qualified images should be determined first based on the usage scenarios specified by the user and the ratings in the historical usage evaluations. This stage mainly screens the formal attributes of the images to ensure that the selected images meet the user's preliminary requirements in terms of basic conditions. Next, based on the preliminary screening, the evaluation features in the historical usage evaluations should be further analyzed, and matching screening at the content level should be performed based on this to ensure that the final recommended images not only meet the user's needs in form but also highly match the user's focus features in terms of content.

[0086] The image recommendation display module is used to interactively display the screened adaptive images. Specifically, the interactive display is implemented as follows: the screened adaptive images are arranged in descending order according to the rating star rating, and the first-ranked adaptive image is extracted from the descending order results and displayed in the search results of the search page, and a navigation button is provided to switch to other adaptive images.

[0087] The present invention ensures that the user sees the most popular picture first by displaying the first-ranked adapted picture in the search results of the search page, thereby improving the user's satisfaction with the recommendation results. At the same time, the user does not have to browse all the pictures one by one, but can directly see the best options after screening and sorting, saving the user's time and energy. In addition, the navigation button provided allows the user to easily browse other adapted pictures, thereby increasing the user's interactive depth. More importantly, the user will have a greater sense of participation in the process of switching pictures using the navigation button, thereby increasing the user's stickiness to the platform. In general, through this interactive display method, the user can easily view the picture ranked first, and switch to other pictures through simple input, thereby achieving an efficient browsing and decision-making process.

[0088] The scene configuration library is used to store background objects that are conventionally configured for different scenes.

[0089] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they shall all fall within the protection scope of the present invention.

Claims

1. An artificial intelligence-based picture trading platform management system, characterized in that: Includes the following modules: The picture display upload module is used to set an upload entry in the picture display page of the trading platform, allowing historical buyers to upload the picture's usage status image and usage evaluation on the picture details page; The usage scenario dynamic update module is used to regularly identify valid positive reviews from the usage status images and usage reviews uploaded on the picture details page according to the set update cycle, and extract usage scenarios from the usage status images of valid positive reviews, and then update the extracted usage scenarios on the picture details page; The image retrieval module is used to provide usage scenario retrieval conditions on the image retrieval page of the trading platform, allowing users to enter usage scenarios for image retrieval; The first picture screening module is used to preliminarily screen out pictures that meet the user's needs from each picture detail page in the picture display page according to the usage scenario input by the user, and record them as candidate pictures, and retrieve the usage evaluation of the candidate pictures, thereby deeply screening out the available pictures; The second picture screening module is used to identify the attention features of the effective positive comments of the available pictures, and then retrieve the user's historical transaction records to analyze the user's tendency attention features, thereby matching the attention features of the available picture usage evaluation with the user's tendency attention features to screen out suitable pictures; The picture recommendation display module is used to interactively display the selected suitable pictures; The scene configuration library is used to store background objects that are commonly configured for different scenes.

2. The image trading platform management system based on artificial intelligence as claimed in claim 1, characterized in that: The identification process of effective positive evaluation is as follows: Mapping the usage evaluation and usage status image of each image uploaded on the image display page within the update cycle; The uploaded usage evaluation is segmented into several words, and the words are tagged with parts of speech; Based on the part of speech of each participle, the evaluation representation phrase is extracted. If the evaluation representation phrase is not extracted from a certain usage evaluation, the usage evaluation is removed, and the remaining usage evaluation is recorded as a valid evaluation; The rating stars are extracted from the valid reviews, and the emotional tendencies corresponding to the valid reviews are obtained based on the corresponding relationship between the rating stars and the emotional tendencies, and then the valid reviews with positive emotional tendencies are screened out as valid positive reviews.

3. The image trading platform management system based on artificial intelligence as claimed in claim 2, characterized in that: The evaluation characterization phrase is extracted as follows: Adjectives and adverbs are collectively referred to as descriptive words, and based on the part of speech of each participle, the participles with the part of speech of noun and descriptive words are selected and recorded as noun participles and descriptive participles respectively; Adjacent noun participles and descriptive word participles are combined into phrases as evaluation representation phrases.

4. The image trading platform management system based on artificial intelligence as claimed in claim 1, characterized in that: The process of extracting the usage scenario from the usage status image with valid positive evaluation is as follows: Performing background separation on the usage status images with effective positive evaluation, and identifying background objects from the separated background; Count the number of recognized background objects, and match each background object with the background objects that are conventionally equipped for different scenes in the scene configuration library, thereby obtaining a scene in which each background object matches; The scenes matching the background objects are classified as the same scenes, the occurrence frequency of each scene is counted, and then the scene corresponding to the maximum occurrence frequency is used as the usage scene for effectively and positively evaluating the corresponding usage status image.

5. The image trading platform management system based on artificial intelligence as claimed in claim 2, characterized in that: The available pictures are screened as follows: Count the percentage of valid positive reviews in the usage reviews uploaded by historical purchasers corresponding to the candidate images, and compare them with the set target percentage, and extract the candidate images that meet the target percentage as reference images; Extract the rating stars from the usage reviews of the reference images, and compare the rating stars of each usage review of each reference image to obtain the median rating star and average rating star of each reference image; The median average deviation of each reference image is obtained by taking the absolute value of the difference between the median rating and the average rating of each reference image and dividing it by the median rating. The median average deviation of each reference image is compared with the system-limited deviation. If the median average deviation of a reference image reaches the system-limited deviation, the reference image is discarded and the remaining reference images are used as available images.

6. The image trading platform management system based on artificial intelligence as claimed in claim 2, characterized in that: The implementation of identifying the focus features for the effective positive evaluation of the available pictures is as follows: Extract effective positive reviews from the usage reviews of available images, and separate the review features and review descriptions from the review representation phrases extracted from the reviews; The extracted evaluation representation phrases are numbered according to the order in which they appear in the usage evaluation, and the numbers are used as the occurrence numbers of the evaluation features obtained by splitting the evaluation representation phrases; Extracting degree words from the evaluation descriptions split from each evaluation representation phrase, and obtaining weight values ​​corresponding to the degree words according to a predefined weight correspondence relationship between the degree words and the sentiment intensity; The weight values ​​corresponding to each degree word in the evaluation description corresponding to each evaluation representation phrase are accumulated to obtain the sentiment intensity coefficient of the evaluation feature corresponding to the evaluation representation phrase; Substitute the occurrence number and sentiment intensity coefficient corresponding to each evaluation feature in the available pictures into the formula Calculate the attention corresponding to each evaluation feature in the available image's effective positive evaluation , where Indicates the occurrence number of the evaluation feature, represents the number of evaluation representation phrases, The coefficient representing the sentiment intensity of the evaluation feature; Classify the evaluation features existing in all valid positive evaluations of the available images according to the same evaluation features, and obtain a number of valid positive evaluations corresponding to each evaluation feature; The attention degree of each evaluation feature in different valid positive evaluations is accumulated and calculated to obtain the total attention degree of each evaluation feature, and then the evaluation feature corresponding to the maximum total attention degree is extracted as the attention feature of the available image.

7. The image trading platform management system based on artificial intelligence as claimed in claim 6, characterized in that: The user's tendency to pay attention to the following features is analyzed: When the user has historical transaction records, the transaction image is extracted from the user's historical transaction records, and the usage evaluation of the transaction image is retrieved from the transaction image details page; Similarly, effective positive comments are screened out from the usage evaluation of the transaction pictures, and the effective positive comments are similarly subjected to attention feature recognition to obtain the attention features of the transaction pictures; The attention features of each historical transaction record and the corresponding transaction image are compared, and the attention features with the highest frequency of occurrence are extracted as the user's preferred attention features.

8. The image trading platform management system based on artificial intelligence as claimed in claim 7, characterized in that: The user tendency attention feature also includes the following analysis process: When the user does not have any historical transaction records, the historical transaction records are retrieved based on the usage scenario input by the user, and then the historical transaction records that meet the usage scenario input by the user are extracted from the retrieval records involved in the historical transaction records and recorded as associated historical transaction records; The associated historical transaction records are similarly analyzed according to the analysis process of the features that the user tends to pay attention to when the user has historical transaction records.

9. The image trading platform management system based on artificial intelligence as claimed in claim 1, characterized in that: The adapted pictures are screened in the following manner: The attention features of available images are matched with the attention features of users, and the available images with successful matching are selected as adapted images.

10. The image trading platform management system based on artificial intelligence as claimed in claim 2, characterized in that: The interaction shows the following implementation process: The selected suitable images are arranged in descending order according to the rating star, and the first-ranked suitable image is extracted from the descending order results and displayed in the search results of the search page, and a navigation button is provided to switch to other suitable images.

Images