Display article auditing method and system based on artificial intelligence

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a displayed article auditing method and system based on artificial intelligence, and the method comprises the steps: carrying out the auditing of an article at each display position through a first standard of a display rule; if the auditing result is not that the display is reasonable, according to the auditing result of the first standard, performing auditing of a second standard on the displayed article; according to the results of the two times of auditing, generating a joint auditing result; and if the result of the joint audit is that the display is unreasonable, generating an adjustment strategy. And the comprehensive judgment capability from the whole to the local and from rule matching to semantic understanding is realized. The display adjustment scheme under the minimum intervention can be effectively searched, and the accuracy and operability of the adjustment strategy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and particularly to an auditing method and system for displayed items based on artificial intelligence. Background Art

[0002] In actual scenarios such as retail, exhibitions, and supermarkets, the display method of commodities not only affects consumers' visual perception and purchase desire, but also is directly related to brand promotion effects and sales efficiency. Therefore, how to effectively, reasonably, and standardly audit the display status of commodities has become one of the key issues in retail intelligent and automated management.

[0003] Currently, traditional display audits mainly rely on manual inspections or simple rule template comparisons, which have many problems such as low audit efficiency, strong subjectivity, inconsistent standards, and inability to adapt to diverse scenarios. With the development of computer vision and artificial intelligence technologies, some enterprises have tried to introduce image recognition models for commodity detection and recognition. However, existing technologies generally stay at the target recognition level, lacking in-depth semantic understanding and structural rule analysis of "display rationality", and it is difficult to meet the business requirements of precise audits.

[0004] Especially when facing complex display rules such as brand characteristics, layout logic, and hierarchical structures, existing static template comparison-based solutions are difficult to support dynamic adaptation and policy feedback, nor can they generate optimal adjustment plans according to scene changes. More importantly, the lack of coordinated judgment on the relationship between local rationality and global structure leads to the audit results being limited to the local area and unable to form an overall optimization guidance. Summary of the Invention

[0005] In view of the above problems, the present invention is proposed.

[0006] To solve the above technical problems, the present invention provides the following technical solution: An auditing method for displayed items based on artificial intelligence, including: Obtain an image of the item display in the display area, and identify the display position of each displayed item according to the image content; Audit the items at each display position using the first standard of the display rules; if the audit result is that the display is reasonable, directly output the audit result; if the audit result is not that the display is reasonable, then audit the displayed items using the second standard according to the audit result of the first standard; Generate a combined audit result based on the results of the two audits; if the combined audit result is that the display is reasonable, directly output the combined audit result; if the combined audit result is that the display is unreasonable, generate an adjustment strategy; The adjustment strategy includes adjusting the item display to make the combined audit result be that the display is reasonable, and outputting the plan with the shortest adjustment path.

[0007] As a preferred solution of the method for auditing display items based on artificial intelligence according to the present invention, wherein: the recognition result of the image content includes identifying the position of the display item, the defect of the display item, and the type of the display item through a trained object detection model; The recognition process of the object detection model is divided into two stages: in the first stage, the object detection model is used to directly identify the position of the display item and the type of the display item; In the second stage, using the type of the display item at each position, the corresponding attention weight is matched through a preset mapping relationship; according to the attention weight, attention allocation is performed on the object detection model; the object detection model after attention allocation is used to separately identify the defects of the display items at each position.

[0008] As a preferred solution of the method for auditing display items based on artificial intelligence according to the present invention, wherein: after the recognition is completed, the defects of each display item are recorded, and while outputting the audit result, the defect recognition result is also output.

[0009] As a preferred solution of the method for auditing display items based on artificial intelligence according to the present invention, wherein: the first standard includes auditing the display according to the consistency principle of the preselected features; According to the preselected features during human-computer interaction before the audit, if the preselected features are non-numerical features, then the non-numerical preselected features are quantified, and the audit is performed according to the quantified preselected features; For each row in the display area, the consistency of the features is audited: the maximum and minimum values of the selected features are screened for each row, so that in one direction or from the middle position to both sides, the minimum value of the features in the previous row is greater than or equal to the maximum value of the features in the current row; and in each row, the feature values of the items decrease in one direction or from the middle position to both sides; When the audit of the feature consistency is satisfied, it is determined that the audit result is that the display is reasonable.

[0010] As a preferred solution of the method for auditing display items based on artificial intelligence according to the present invention, wherein: the second standard includes inheriting the preselected features in the first standard to perform a rationality audit of the display; wherein, the rationality is classified into: regional rationality and supplementary rationality; The supplementary rationality is completed through quantitative analysis. When the preselected features are non-numerical features, in the nth row, if the number of remaining display positions is equal to the number of items with feature j after the items with feature i are displayed, it is determined that the nth row meets the rationality; otherwise, it is determined that the nth row does not meet the rationality; Wherein, i and j represent different feature indices in the same kind of feature; The regional rationality is achieved by dividing the display area: in the display area, sub-areas are divided to obtain a combination of sub-areas, so that the display of items in each sub-area meets the review of the said feature consistency or the comprehensive review of the said feature consistency and the supplementary rationality; the combination of sub-areas with the fewest sub-areas is output as the result of the division of the display area. Through a pre-trained neural network, a binary classification judgment on whether the result of the regional division is reasonable is obtained by analyzing the rationality of the result of the division of the display area.

[0011] As a preferred solution of the method for reviewing displayed items based on artificial intelligence according to the present invention, wherein: the joint review includes, first, comprehensively reviewing the feature consistency and the supplementary rationality according to the regional rationality. If the result of the comprehensive review is that the display is unreasonable, then the regional rationality is judged. If the judgment result of the regional rationality is that the result of the regional division is reasonable, then the result of the joint review is output as the display being qualified; otherwise, the result of the joint review is output as the display being unqualified. The comprehensive review includes, during the judgment process of the regional rationality, removing all rows judged to be reasonable in the display area. If the review result of the feature consistency in the display area after removal is that the display is reasonable, then the result of the joint review is directly output as the display being qualified; otherwise, the result of the comprehensive review is generated as the display being unreasonable.

[0012] As a preferred solution of the method for reviewing displayed items based on artificial intelligence according to the present invention, wherein: the adjustment strategy further includes using an optimization algorithm based on energy conservation to generate a strategy. The optimization algorithm based on energy conservation includes assuming that each displayed item is an indivisible energy; in the display area, by rearranging the energy, the global state is made stable. Assume that the display of items judged to be reasonable through review represents the global state being stable. In the display area, when removing displayed items at any position, after removing D displayed items, if the global state is stable, then record the positions and item marks of the D displayed items; by testing different removal strategies, select the positions and marks of the displayed items corresponding to the minimum value of D. Denoted as: the minimum value of the removal quantity The set of positions of the corresponding displayed items The marks of the corresponding displayed items ; wherein, represents the r-th removal display position, represents the r-th removed item mark; Indicates the number of items removed; The specific steps of the rearrangement are as follows: Step 1: Incorporate the empty display positions into W to obtain , which serves as the positions available for arrangement; Step 2: Randomly arrange the elements in Q within , and analyze whether the overall situation is in a steady state after the arrangement; Step 3: If there is an arrangement plan in a steady state, calculate the sum L of the distances of each item before and after adjustment, and output the arrangement plan with the minimum L; if there is no arrangement plan in a steady state, randomly select one of the non-removed display items as the removed display item, and re-analyze Steps 1 and 2. After visiting all the removed display items, output the arrangement plan with the minimum L; when randomly selecting one of the non-removed display items still cannot obtain an output, increase the randomly selected quantity until an output is obtained.

[0013] An audit system for display items based on artificial intelligence, wherein: A collection unit that acquires images of item displays in the display area and identifies the display positions of each display item according to the image content; An audit unit that audits the items at each display position using the first standard of display rules; if the audit result is that the display is reasonable, directly output the audit result; if the audit result is not that the display is reasonable, then conduct a second standard audit on the displayed items according to the audit result of the first standard; An analysis unit that generates a combined audit result based on the results of the two audits; if the combined audit result is that the display is reasonable, directly output the combined audit result; if the combined audit result is that the display is unreasonable, generate an adjustment strategy.

[0014] A computer device, including: a memory and a processor; the memory stores a computer program, wherein: when the processor executes the computer program, the steps of the method described in any one of the present inventions are implemented.

[0015] A computer-readable storage medium, on which a computer program is stored, wherein: when the computer program is executed by a processor, the steps of the method described in any one of the present inventions are implemented.

[0016] Advantages of the present invention: The method for auditing displayed items based on artificial intelligence provided by the present invention realizes multi-dimensional intelligent auditing of displayed items by introducing multi-stage object detection, pre-selection feature consistency auditing mechanism, and regional and supplementary rationality judgment strategies. Through structure analysis based on image recognition and combined with auditing features set by human-computer interaction, it realizes comprehensive judgment ability from the whole to the part, from rule matching to semantic understanding. In particular, the present invention adopts an optimization algorithm based on the idea of energy conservation, which can effectively find a display adjustment plan with the least intervention and improve the accuracy and operability of the adjustment strategy. Compared with the prior art, the present invention has significantly improved in terms of auditing accuracy, adaptability to diverse scenarios, and automatic adjustment ability, and is applicable to various scenarios such as large supermarkets, chain brands, and intelligent stores, with high practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0018] Figure 1 It is the overall flowchart of a method for auditing displayed items based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will give a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0020] Refer to Figure 1 , for an embodiment of the present invention, a method for auditing displayed items based on artificial intelligence is provided, including: S1: Obtain an image of the item display in the display area, and identify the display position of each displayed item according to the image content.

[0021] Further, through a trained object detection model, the identification of the position of the displayed item, the defect of the displayed item, and the type of the displayed item is performed. The identification process of the object detection model is divided into two stages: In the first stage, the object detection model is used to directly identify the position of the displayed item and the type of the displayed item.

[0022] In the second stage, using the types of display items at each position, through a preset mapping relationship, the corresponding attention weights are matched (these weights are preset, generally for the positions where a certain item is prone to defects, and attention weighting is performed); according to the attention weights, attention allocation is performed on the target detection model; the target detection model after attention allocation is used to respectively identify the defects of the display items at each position.

[0023] A two-stage recognition structure is introduced: the first stage focuses on macroscopic recognition, quickly locks the position and type information of each item, and establishes a basic structure map; the second stage introduces the semantic association between the type and the defect, and through a preset mapping mechanism, matches the identified item type with the corresponding attention weight to guide the target detection model to focus on the areas prone to defects, thereby improving the accuracy and robustness of defect detection.

[0024] It can adaptively adjust the model's attention area according to the characteristic differences of different types of items, realize type-driven attention allocation and depth perception optimization, thereby effectively improving the performance of defect recognition in complex display environments. At the same time, the staged recognition also realizes the dynamic scheduling of model resources, improves the recognition efficiency, and is suitable for the high-concurrency and high-precision audit requirements in actual retail scenarios.

[0025] In this embodiment, the target detection model is preferably a YOLOv5 (You Only Look Once version 5) model constructed based on a deep convolutional neural network. This model has end-to-end object localization and classification capabilities, and can achieve high-precision detection of multiple commodity targets in the display image while ensuring real-time performance. The YOLOv5 model extracts features through the CSPDarknet backbone network and combines the PANet structure to achieve cross-layer feature fusion, and can effectively identify the position bounding box, type label and confidence value of the item, meeting the recognition requirements for multi-target and densely arranged targets in the retail scenario.

[0026] In the first-stage recognition process, the YOLOv5 model is used to obtain the bounding box position and category identifier of each item, providing input conditions for the attention guidance in the second stage. This model has completed transfer training on a display dataset containing images of multiple types of commodities and has strong scene adaptation capabilities.

[0027] In an alternative embodiment, the target detection model may also include, but is not limited to: models such as YOLOv8, Faster R-CNN, DETR, etc.

[0028] After the recognition is completed, the defects of each display item are recorded, and while outputting the audit result, the defect recognition result is also output.

[0029] S2: Use the first criterion of the display rule to review the items at each display position. If the review result shows reasonable display, directly output the review result. If the review result does not show reasonable display, then according to the review result of the first criterion, conduct a review of the items on display using the second criterion.

[0030] The first criterion includes conducting a review of the display according to the consistency principle of the preselected features.

[0031] Based on the features preselected during human-computer interaction before the review (for example, using features such as the volume, height, brand, year, or usage of a preset item as criteria. When presetting, multiple features can be set), if the preselected features are non-numerical features, then quantify the non-numerical preselected features and conduct a review according to the quantified preselected features (for example, if the preset feature is color, then first quantify the color feature and then conduct an analysis).

[0032] For each row in the display area, conduct a review of the feature consistency: screen the maximum and minimum feature values for each row, so that in one direction or from the middle position to both sides, the minimum feature value of the upper row is greater than or equal to the maximum feature value of the current row. And in each row, the feature values of the items decrease in one direction or from the middle position to both sides. In fact, it can be simply understood as: when conducting a review of the display feature consistency, for each row in the display area, based on the preselected feature set extract the set of corresponding feature values of all items in that row where indicates that the current row is the th row; define the maximum and minimum feature values of the th row as: and .

[0033] If there are two rows that satisfy the following decreasing relationship: . Then it is considered that the judgment condition of "hierarchical feature decreasing consistency" is met between the current two rows.

[0034] If the review mode is set to "symmetric decreasing from the center to both sides", then set the current row as row, and the column index range is ; if , then ; otherwise, ; that is, the items decrease from the middle column to the left and right sides, forming a symmetric feature distribution.

[0035] Based on the consistency principle of preselected features, a basic display audit rule system with both universality and structured judgment logic is constructed. In actual retail, display, warehousing and other scenarios, item display has significant hierarchical, symmetrical and feature classification requirements. This design preselects audit features through human-computer interaction, such as brand, volume, color, use, etc., to achieve personalized and adaptive audit logic configuration.

[0036] To support a wider range of audit feature adaptation, this solution specially introduces a quantification mechanism for non-numerical features. For example, color is encoded as an HSV value, and brand is mapped to a rank number, etc., so that unstructured information can be converted into a unified numerical scale, facilitating algorithm processing and consistency calculation.

[0037] This design further introduces a decreasing consistency rule between rows at the structural level, requiring that the minimum feature value of the previous row should not be less than the maximum value of the current row, strengthening the hierarchical progression logic of the display. At the same time, a layout rule of "central symmetry decreasing" or "unidirectional decreasing" is introduced in the horizontal structure, enabling the system to have stronger symmetry recognition ability when detecting the internal structure of each row. This audit mechanism is applicable not only to numerical features with natural order relationships such as height and size, but also to feature values with "soft sorting" logic such as hue gradient and category aggregation, greatly improving the applicability and fault tolerance of the audit algorithm. When the audit of the feature consistency is satisfied, the audit result is judged as reasonable display.

[0038] The second standard includes inheriting the preselected features in the first standard to conduct a rationality audit of the display; among them, the classification of rationality is: regional rationality and supplementary rationality.

[0039] The supplementary rationality is completed through quantitative analysis. When the preselected feature is a non-numerical feature, in the nth row, if the number of remaining display positions is equal to the number of items of feature j after the items of feature i are displayed, it is judged that the nth row meets the rationality; otherwise, it is judged that the nth row does not meet the rationality. This judgment refers to "items with a small quantity" making "void supplements" in the display. If the quantity of such items is too large, it means that the supplementary items actually form a scale and are not suitable for supplementing the display positions of items with a small base number. Among them, i and j represent different feature indexes in the same kind of feature.

[0040] Traditional consistency audit methods tend to judge the arrangement logic and unity of overall features. However, for individual categories with a small quantity in the store that still need to be displayed, if forced to be restricted by unified rules, it may lead to misjudgment of the audit or even misidentification of reasonable displays as unqualified.

[0041] For this reason, a supplementary rationality judgment mechanism is proposed. Taking the preselected features as the reference basis, non-numerical features are quantified and encoded, and the feature distribution of items in the same row is analyzed structurally. Specifically, if the number of remaining empty spaces in the current row is exactly equal to the number of another type of small-cardinality feature (feature j) after the main display of a certain type of feature item (denoted as feature i), then feature j is considered a reasonable supplementary item, and its existence constitutes a natural extension of the main display logic. This judgment logic reflects the design principle of "small quantity, filling, and not forming a dominant structure", aiming to achieve automatic judgment through structural analysis and quantity matching, identifying supplementary items that will not disrupt the main logical display order visually and structurally, and enhancing the system's flexible adaptability to complex actual scenarios.

[0042] Conversely, if the number of items of a certain feature j exceeds the remaining empty spaces, forming a main structural occupancy, it indicates that this category no longer has the "supplementary attribute" and should not be considered a reasonable supplementary position, thus judging that the display in this row is unreasonable. This mechanism takes into account the proportional relationship and structural rationality of the feature dimension, effectively improving the inclusiveness and intelligence of the audit strategy, and is particularly suitable for complex layouts such as non-standard structures, temporary displays, and combined sales scenarios.

[0043] The regional rationality is achieved by dividing the display area: in the display area, sub-areas are divided to obtain a sub-area combination, so that the item display in each sub-area meets the audit of the feature consistency or the comprehensive audit of the feature consistency and the supplementary rationality (this condition refers to any of the following: 1. Each area meets the consistency; 2. The consistency after being supplemented by "supplementary rationality"); the sub-area combination with the fewest sub-areas is output as the result of the division of the display area. Through a pre-trained neural network, a binary classification judgment on whether the result of the division of the display area is reasonable is obtained for the result of the division of the display area.

[0044] S3: Generate the result of the joint audit based on the results of the two audits.

[0045] Specifically, it includes: first, comprehensively auditing the feature consistency and the supplementary rationality according to the regional rationality. If the result of the comprehensive audit shows that the display is unreasonable, then judge the regional rationality. If the judgment result of the regional rationality is that the result of the regional division is reasonable, then output the result of the joint audit as the display being qualified; otherwise, output the result of the joint audit as the display being unqualified.

[0046] The comprehensive review includes, during the judgment of regional rationality, removing all rows judged to be reasonable from the display area. If, after the removal, the review result of the feature consistency for the display area is that the display is reasonable, then directly output the result of the joint review as qualified display; otherwise, generate the result of the comprehensive review as unreasonable display.

[0047] It should be noted that the "regional rationality review mechanism" proposed in this solution aims to solve the situation where the overall consistency review fails in actual display reviews due to the diversification of item features and the complexity of structures. During the standard review process, once the features of a certain part of the items are inconsistent, it is easy to judge the entire row or entire area as "unreasonable", while ignoring the reasonable display units with good structures within the local area.

[0048] In this embodiment, the neural network used for the rationality analysis of the display area segmentation result is preferably a convolutional neural network (CNN, Convolutional Neural Network) structure. This neural network performs a binary classification judgment on whether the segmentation result conforms to visual and structural rationality by inputting the combined image of the divided sub-areas or its corresponding structured feature matrix. The CNN model is pre-trained on a dataset containing a large number of display images and their corresponding rationality labels, and has strong feature extraction and classification discrimination capabilities, and can accurately identify whether there are problems such as structural damage, feature conflicts, or boundary misalignments in the area combination, thereby providing an effective judgment basis for subsequent reviews.

[0049] In an alternative embodiment, the above neural network can also be any one or a combination of the following structures: multi-layer perceptron (MLP), graph neural network (GNN), attention mechanism network (such as Transformer structure), fusion model (CNN+LSTM or CNN+Transformer), etc.

[0050] S4: If the result of the joint review is qualified display, then directly output the result of the joint review; if the result of the joint review is unreasonable display, then generate an adjustment strategy.

[0051] The adjustment strategy includes adjusting the item display to make the result of the joint review be qualified display, and outputting the plan with the shortest adjustment path. Using an optimization algorithm based on energy conservation to generate the strategy: The optimization algorithm based on energy conservation includes assuming that each displayed item is an indivisible energy; in the display area, by rearranging the energy, the global state is made stable. Assuming that the item display judged to be reasonable through the review represents a stable global state.

[0052] In the display area, at any position, remove the displayed items. After removing D displayed items, if the global state is stable, record the positions and item tags of the D displayed items; by testing different removal strategies, select the positions and tags of the displayed items corresponding to the minimum value of D.

[0053] Denoted as: the minimum value of the removal quantity , the set of positions of the corresponding displayed items , the tags of the corresponding displayed items .

[0054] Among them, represents the r-th removed display position, represents the r-th removed item tag; represents the number of removed items.

[0055] The specific steps of the rearrangement are as follows: Step 1: Incorporate the vacant display positions into W to obtain , as the positions available for arrangement.

[0056] Step 2: Randomly arrange the elements in Q in (each arrangement method is an arrangement plan), and analyze whether the global state is stable after the arrangement.

[0057] Step 3: If there is an arrangement plan in a stable state, calculate the sum L of the distances of each item before and after adjustment, and output the arrangement plan with the minimum L; if there is no arrangement plan in a stable state, randomly select one of the non-removed displayed items as the removed displayed item, and re-analyze Steps 1 and 2. After traversing all the removed displayed items, output the arrangement plan with the minimum L.

[0058] When randomly selecting one of the non-removed displayed items still cannot obtain an output, increment the randomly selected quantity (after each increment of the quantity, traverse all the plans; if still unable to output a plan, then increment the quantity again) until a plan is output.

[0059] The "optimization algorithm based on energy conservation" proposed in this solution aims to solve the problem that unreasonable structures are found during the display review but cannot be automatically adjusted by simple rules. Traditional adjustment strategies usually rely on fixed templates or manual experience and are difficult to adapt to the complex adjustment requirements brought about by category mixing, uneven commodity quantities, and tight layouts in the display area. Especially in the case where the goal is minimum intervention and optimal layout restoration, there is a lack of a mathematical and systematic strategy generation mechanism.

[0060] To this end, the present invention introduces the idea of "energy conservation" for optimized modeling. Each displayed item is regarded as an indivisible "energy unit", and the currently approved reasonable state is defined as the "system steady state". The display area is abstracted into a system with balanced energy distribution. By removing some energy (i.e., a small number of items) and redistributing the remaining energy in the empty spaces, a new steady-state structure is sought to achieve an optimized transition from an unreasonable state to a reasonable state.

[0061] The core objective of the design of this optimization algorithm is to find the minimum removal set D to restore the structural rationality and ensure that the adjustment plan has the least interference with the original display. Based on combinatorial search, candidate layout plans are constructed, and the optimal solution is explored by traversing the removal and rearrangement paths. Using the sum of the item movement distances L as the evaluation index, the path with the minimum overall displacement is selected, taking into account both rationality and adjustment cost. An iterative increment strategy is introduced to automatically expand the removal range when the initial removal fails, ensuring the outputability of the final solution. The algorithm converges to the minimum adjustment path, making the adjustment strategy interpretable, stable, and operable.

[0062] Through this strategy design, the system can automatically generate an adjustment plan that meets the global structural steady state without relying on a hard rule template, and can adapt to multi-dimensional characteristics such as product types, levels, and positions, greatly improving the flexibility, intelligence, and engineering practicality of the adjustment logic.

[0063] On the other hand, the present embodiment also provides an audit system for displayed items based on artificial intelligence, which includes: A collection unit that acquires the images of item displays in the display area and identifies the display positions of each displayed item according to the image content.

[0064] An audit unit that audits the items at each display position using the first standard of the display rules; if the audit result is that the display is reasonable, the audit result is directly output; if the audit result is that the display is not reasonable, the items on display are audited using the second standard according to the audit result of the first standard.

[0065] An analysis unit that generates a combined audit result based on the results of the two audits; if the combined audit result is that the display is reasonable, the combined audit result is directly output; if the combined audit result is that the display is unreasonable, an adjustment strategy is generated.

[0066] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0067] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

[0068] More specific examples (non-exhaustive list) of computer-readable media include the following: electrical connection parts with one or more wirings (electronic devices), portable computer disk cartridges (magnetic devices), random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memories), fiber optic devices, and portable compact disc read-only memories (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as necessary, and then storing it in a computer memory.

[0069] It should be understood that each part of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0070] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based method for auditing displayed items, characterized in that, Including: Obtain the image of item display in the display area, and identify the display position of each displayed item according to the image content; Use the first criterion of the display rule to review the items at each display position; if the review result is that the display is reasonable, directly output the review result; If the review result is not that the display is reasonable, then according to the review result of the first criterion, conduct a second criterion review on the displayed items; Generate a combined review result based on the results of the two reviews; if the combined review result is that the display is reasonable, directly output the combined review result; if the combined review result is that the display is unreasonable, generate an adjustment strategy; The adjustment strategy includes making the combined review result be that the display is reasonable by adjusting the item display, and outputting the plan with the shortest adjustment path.

2. The method for auditing displayed items based on artificial intelligence according to claim 1, wherein: The recognition result of the image content includes identifying the position of the displayed item, the defect of the displayed item, and the type of the displayed item through a trained object detection model; The recognition process of the object detection model is divided into two stages: in the first stage, directly identify the position of the displayed item and the type of the displayed item by using the object detection model; In the second stage, use the type of the displayed item at each position to match the corresponding attention weight through a preset mapping relationship; Conduct attention allocation on the object detection model according to the attention weight; Use the object detection model after attention allocation to separately identify the defects of the displayed items at each position.

3. The method for auditing displayed items based on artificial intelligence according to claim 2, wherein: After the recognition is completed, record the defects of each displayed item, and output the defect recognition result while outputting the review result.

4. The method for auditing display items based on artificial intelligence according to claim 3, characterized in that: The first criterion includes conducting a review of the display according to the consistency principle of the preselected features; According to the preselected features during human-computer interaction before the review, if the preselected feature is a non-numerical feature, then quantify the non-numerical preselected feature and conduct a review according to the quantified preselected feature; Conduct a review of feature consistency for each row in the display area: screen the maximum and minimum feature values for each row, so that in one direction or from the middle position to both sides, the minimum feature value of the previous row is greater than or equal to the maximum feature value of the current row; and in each row, the feature values of the items decrease in one direction or from the middle position to both sides; When the review of the feature consistency is satisfied, judge that the review result is that the display is reasonable.

5. The method for auditing display items based on artificial intelligence according to claim 4, wherein: The second criterion includes inheriting the preselected features in the first criterion to conduct a review of the display rationality; among them, the classification of rationality is: regional rationality and supplementary rationality; The supplementary rationality is completed through quantitative analysis. When the preselected feature is a non-numerical feature, in the nth row, if the number of remaining display positions is equal to the number of items with feature j after the items with feature i are displayed, then judge that the nth row meets the rationality; otherwise, judge that the nth row does not meet the rationality; Among them, i and j represent different feature indexes in the same type of feature; The regional rationality is achieved by dividing the display area: in the display area, sub-areas are divided to obtain a combination of sub-areas, so that the item display in each sub-area meets the review of the feature consistency or the comprehensive review of the feature consistency and the supplementary rationality; the sub-area combination with the fewest sub-areas is output as the result of the division of the display area. Through a pre-trained neural network, a binary classification judgment on whether the result of the division of the display area is reasonable is obtained for the rationality analysis of the result of the division of the display area.

6. The method for auditing display items based on artificial intelligence according to claim 5, characterized in that: The joint review includes, first, comprehensively reviewing the feature consistency and the supplementary rationality according to the regional rationality. If the result of the comprehensive review is that the display is unreasonable, then the regional rationality is judged. If the judgment result of the regional rationality is that the result of the regional division is reasonable, then the result of the joint review is output as the display being qualified; otherwise, the result of the joint review is output as the display being unqualified. The comprehensive review includes, during the judgment process of the regional rationality, all rows judged to be reasonable are removed from the display area. If the review result of the feature consistency in the remaining display area is that the display is reasonable, then the result of the joint review is directly output as the display being qualified; otherwise, the result of the comprehensive review is generated as the display being unreasonable.

7. The method for auditing displayed items based on artificial intelligence according to claim 6, wherein: The adjustment strategy also includes generating a strategy using an optimization algorithm based on energy conservation. The optimization algorithm based on energy conservation includes assuming that each displayed item is an indivisible energy. In the display area, by rearranging the energy, the global state is made stable. Assume that the item display judged to be reasonable through the review indicates that the global state is stable. In the display area, when removing displayed items at any position, after removing D displayed items, if the global state is stable, record the positions and item marks of the D displayed items; by testing different removal strategies, select the positions and marks of the displayed items corresponding to the minimum value of D. Denoted as: the minimum value of the elimination quantity , the position set of the corresponding display items , the label of the corresponding display item ; Among them, represents the r-th removed display position, represents the r-th removed item label; represents the number of removed items; The specific steps of the rearrangement are as follows: Step 1: Incorporate the vacant display positions into W to obtain , which serves as the available layout positions; Step 2: Randomly arrange the elements in Q in and analyze whether the whole is in a steady state after the arrangement; Step 3: If there is a stable arrangement plan, calculate the sum L of the distances of each item before and after adjustment, and output the arrangement plan with the minimum L; if there is no stable arrangement plan, randomly select one of the non-removed displayed items as the removed displayed item, and re-analyze Steps 1 and 2. After traversing all the removed displayed items, output the arrangement plan with the minimum L; when randomly selecting one still cannot obtain the output among the non-removed displayed items, increase the number of random selections until the output is obtained.

8. An artificial intelligence-based display item review system using the method according to any one of claims 1-7, characterized in that: The acquisition unit obtains the image of the item display in the display area and identifies the display position of each displayed item according to the image content. The review unit reviews the items at each display position using the first standard of the display rules; if the review result is that the display is reasonable, directly output the review result. If the review result is not that the display is reasonable, then review the displayed items using the second standard according to the review result of the first standard. An analysis unit generates a combined audit result based on the results of two audits; if the combined audit result is that the display is reasonable, the combined audit result is directly output; if the combined audit result is that the display is unreasonable, an adjustment strategy is generated.

9. A computer device, comprising: A memory and a processor; the memory stores a computer program, characterized in that: when the processor executes the computer program, the steps of the method according to any one of claims 1-7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the method according to any one of claims 1-7 are implemented.

Images