Goods shelf commodity management method and related equipment

Abstract

The invention discloses a goods shelf commodity management method and related equipment, and the method comprises the steps: obtaining a goods shelf picture through an equipment end interface, and obtaining a goods shelf detection image according to the pre-defined recognition region configuration; inputting the shelf detection image into a trained commodity recognition model, and performing single forward propagation and non-maximum suppression processing by using the commodity recognition model to generate a commodity recognition result; and matching the commodity identification result with a commodity information list in an application end database, and updating the commodity information list according to a matching result. The method has higher accuracy, reliability and management efficiency, can effectively solve the problems in the existing scheme, and provides a new solution for the field of shelf management.

Description

Technical Field

[0001] The present application relates to the field of intelligent management, and more specifically, to a shelf merchandise management method and related equipment. Background Art

[0002] In the increasingly fierce market competition, the traditional manual management of shelf goods can no longer meet the modern retail industry's needs for efficiency, accuracy, and personalization. Research on intelligent management of shelf goods has been carried out one after another. The more common management solution in the field of shelf management today is to use image recognition technology to compare real-time shelf images with reference images to determine whether the shelf is out of stock and whether the product placement is accurate. This solution has some obvious defects in practical applications.

[0003] First, when the shelf placement requirements change, existing solutions often cannot automatically collect and update the appropriate reference map, which may lead to system misjudgment and affect the accuracy of shelf management. Secondly, when identifying out-of-stock and placement issues, due to the limitations of image recognition technology, there is a certain error rate, which may affect the efficiency of shelf management. Finally, during the recognition process, factors such as salesperson occlusion and debris on the shelf may lead to system misjudgment, further reducing the reliability of shelf management.

[0004] Based on this, the present application provides a shelf merchandise management solution with higher accuracy, reliability and management efficiency, which can effectively solve the problems existing in the existing solutions and provide a new solution for the shelf management field. Summary of the invention

[0005] In view of this, the present application provides a shelf merchandise management method and related equipment, which have higher accuracy, reliability and management efficiency, can effectively solve the problems existing in the existing solutions, and provide a new solution for the shelf management field.

[0006] A shelf commodity management method, comprising:

[0007] Obtain shelf images through the device-side interface, and obtain shelf detection images based on pre-defined recognition area configurations;

[0008] Inputting the shelf detection image into the trained commodity recognition model, performing a single forward propagation and non-maximum suppression process using the commodity recognition model, and generating a commodity recognition result;

[0009] The commodity identification result is matched with the commodity information list in the application-side database, and the commodity information list is updated according to the matching result.

[0010] Optionally, the commodity identification model includes a format conversion network, a forward propagation network, and a deduplication detection network;

[0011] The format conversion network is used to parse the input shelf detection image into binary image data, and generate an image matrix based on the binary image data conversion;

[0012] The forward propagation network performs commodity detection according to the image matrix through a single forward propagation to obtain a commodity detection result;

[0013] The deduplication detection network performs non-maximum suppression processing on the commodity detection result to generate a commodity recognition result.

[0014] Optionally, matching the commodity identification result with a commodity information list in an application-side database, and updating the commodity information list according to the matching result, includes:

[0015] Match the product recognition result with the product information list in the application-side database in terms of shelf coordinate dimension, product area dimension, and product category dimension respectively;

[0016] If the matching degree of each dimension meets the preset deviation, the product information list is updated according to the product recognition result;

[0017] If there is a target matching dimension that does not meet the preset deviation, the target product information is analyzed based on the target matching dimension, and the target product corresponding to the target product information is determined as a misplaced product.

[0018] Optionally, the product information list also includes a minimum quantity for each category;

[0019] Matching the commodity identification result with the commodity information list in the application-side database also includes:

[0020] The commodity quantity of each commodity in the commodity identification result is matched with the minimum quantity of each category recorded in the commodity information list, out-of-stock commodities whose quantity is lower than the minimum quantity of the corresponding category are confirmed, and replenishment reminders are issued for the out-of-stock commodities.

[0021] Optionally, also include:

[0022] Collect statistics on the identification results of each product within the predetermined analysis period;

[0023] generating a first adjusted commodity set according to the quantity fluctuation of each category of commodities in the commodity identification results within the predetermined analysis period and in combination with the commodity sales volume, wherein the first adjusted commodity set records each commodity whose quantity fluctuation frequency and sales volume ratio is greater than a first predetermined value;

[0024] A product promotion plan is designed based on the first adjusted product set.

[0025] Optionally, also include:

[0026] Collect statistics on the identification results of each product within the predetermined analysis period;

[0027] Generate a second adjusted product set according to the quantity fluctuation of each category of products in the product identification results within the predetermined analysis period, combined with product sales volume and product location information, wherein the second adjusted product set records each product whose quantity fluctuation frequency to sales volume ratio is less than a second predetermined value and whose product location belongs to a non-focus area;

[0028] A merchandise placement plan is designed based on the second adjusted merchandise set.

[0029] A shelf commodity management device, comprising:

[0030] An image acquisition unit, used to acquire a shelf image through a device-side interface and obtain a shelf detection image according to a predefined recognition area configuration;

[0031] A commodity recognition unit, used for inputting the shelf detection image into a trained commodity recognition model, performing a single forward propagation and non-maximum suppression process using the commodity recognition model, and generating a commodity recognition result;

[0032] The matching and updating unit is used to match the commodity identification result with the commodity information list in the application-side database, and update the commodity information list according to the matching result.

[0033] A shelf commodity management device includes a memory and a processor;

[0034] The memory is used to store programs;

[0035] The processor is used to execute the program to implement each step of the shelf commodity management method as described in any one of the above items.

[0036] A readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, each step of the shelf commodity management method as described in any one of the above items is implemented.

[0037] A computer program product includes a computer program, wherein the computer program executes each step of any of the above-mentioned shelf commodity management methods when the computer program is executed by a processor.

[0038] It can be seen from the above technical solutions that a shelf commodity management method and related equipment provided by the embodiment of the present application first obtains a shelf image through a device-side interface, and obtains a shelf detection map according to a predefined recognition area configuration. This step solves the problem of commodity offset, ensures that the system can accurately obtain the commodity area to be identified, and improves the accuracy of recognition. After that, the shelf detection map is input into the trained commodity recognition model, and the commodity recognition model is used to perform a single forward propagation and non-maximum suppression process to generate a commodity recognition result. The commodity recognition model can efficiently identify the types and quantities of commodities on the shelf. Finally, the commodity recognition result is matched with the commodity information list in the application-side database, and the commodity information list is updated according to the matching result. This step realizes real-time monitoring and management of shelf commodities, and can promptly discover problems such as out-of-stock and placement errors. The present application has higher accuracy, reliability and management efficiency, can effectively solve the problems existing in the existing scheme, and provides a new solution for the shelf management field. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0040] Figure 1 A flowchart of a shelf merchandise management method disclosed in an embodiment of the present application;

[0041] Figure 2 A schematic diagram of the structure of a commodity identification model disclosed in an embodiment of the present application;

[0042] Figure 3 A schematic diagram of a shelf commodity management device disclosed in an embodiment of the present application;

[0043] Figure 4 This is a hardware structure block diagram of a shelf merchandise management device disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0045] The present application can be used in many general or special computing device environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multi-processor devices, distributed computing environments including any of the above devices or equipment, etc.

[0046] An embodiment of the present application provides a shelf merchandise management method, which can be applied to shelf merchandise management platforms or applications in offline shopping places such as convenience stores and supermarkets, and can also be applied to various computer terminals or smart terminals. The execution entity can be a processor or server of the computer terminal or smart terminal.

[0047] Next, the present application scheme is introduced. The present application proposes the following technical scheme, please see below for details.

[0048] Figure 1 This is a flow chart of a shelf merchandise management method disclosed in an embodiment of the present application.

[0049] like Figure 1 As shown, the method may include:

[0050] Step S1, obtain a shelf image through a device-side interface, and obtain a shelf detection image according to a predefined recognition area configuration.

[0051] Specifically, through the device-side interface, such as a camera or other image acquisition device, the current image of the shelf is captured in real time, and then finely processed according to the pre-defined recognition area configuration. The configuration of the recognition area is carefully designed based on factors such as shelf layout, product type and placement rules, to ensure that the system can accurately focus on key areas and eliminate unnecessary interference information. Through this configuration, the system can automatically generate a shelf detection map, which not only clearly shows the layout of the goods on the shelf, but also highlights the areas of the goods that need to be identified. Within the identified area, if the shelf is offset, it will not be identified based on the specific coordinates of the product coordinates, but will be mainly based on the dimensions of the regional coordinate range. In this way, within this area, the system misjudgment or inaccurate detection caused by the offset of the overall shelf can be avoided to a certain extent.

[0052] Step S2: input the shelf detection image into the trained commodity recognition model, and use the commodity recognition model to perform single forward propagation and non-maximum suppression processing to generate a commodity recognition result.

[0053] Specifically, the preliminarily processed shelf detection image is input into the pre-trained product recognition model for accurate product recognition. Specifically, this step first passes the shelf detection image as input data to the product recognition model trained by the deep learning algorithm. The model uses a convolutional neural network to perform feature extraction and pattern recognition on the input image, that is, a single forward propagation process. During the forward propagation process, the model processes the input image layer by layer to extract key features that can distinguish different products. Subsequently, the non-maximum suppression (NMS) algorithm is used to further optimize these features. The main purpose of the non-maximum suppression algorithm is to retain the box that is most likely to represent the real product among the multiple candidate boxes detected, while suppressing those boxes with high overlap and low confidence, thereby reducing the possibility of repeated detection and false detection.

[0054] Through this series of complex processing, the product recognition model can generate product recognition results that include product category (sku), location (corresponding regional coordinates), confidence level, etc. These results not only provide key data for subsequent product information matching and updating, but also provide strong support for real-time inventory management of smart shelves, product promotion plans, and product placement plan adjustment and design.

[0055] The product identification model is further introduced below.

[0056] like Figure 2 As shown, the commodity identification model includes a format conversion network 110 , a forward propagation network 120 , and a deduplication detection network 130 .

[0057] The format conversion network 110 is used to parse the input shelf detection image into binary image data, and generate an image matrix based on the binary image data conversion;

[0058] The forward propagation network 120 performs commodity detection according to the image matrix through a single forward propagation to obtain a commodity detection result;

[0059] The deduplication detection network 130 performs non-maximum suppression processing on the commodity detection result to generate a commodity recognition result.

[0060] Specifically, the format conversion network 110 is the starting point of the commodity identification process, and it is responsible for preliminary processing of the input shelf detection image. The format conversion network will parse the shelf detection image into binary image data. This process usually involves binarization of the image, that is, converting each pixel in the image to 0 or 1 to simplify the image information and highlight key features. Subsequently, based on these binary image data, the format conversion network will further convert and generate an image matrix. This matrix is ​​the basis for subsequent processing. It contains the information of all pixels in the image and is presented in a form that is easier for computers to process.

[0061] The forward propagation network 120 receives the image matrix generated by the format conversion network and starts to perform commodity detection to obtain commodity detection results. The forward propagation network usually includes multiple convolutional layers, pooling layers, and fully connected layer structures to extract feature information in the image and perform pattern recognition. Through a single forward propagation, the forward propagation network can analyze the features in the image matrix and generate commodity detection results. These results usually include the location, size, and preliminary classification information of the commodity.

[0062] The deduplication detection network 130 further optimizes the commodity detection results generated by the forward propagation network. In actual scenarios, commodities may overlap or be detected repeatedly due to problems such as dense placement or shooting angles. Therefore, the deduplication detection network uses a non-maximum suppression (NMS) algorithm to filter and deduplicate multiple candidate frames in the detection results. This process retains the frames that are most likely to represent the real commodity and suppresses those frames with high overlap and low confidence, thereby generating the final commodity recognition result.

[0063] Step S3: Match the product identification result with the product information list in the application-side database, and update the product information list according to the matching result.

[0064] Specifically, the product information in the product recognition results, such as product category (SKU), name, location, and quantity, is first compared one by one with the product information list in the database. This process uses efficient database query and matching algorithms to ensure that a large amount of data can be compared in a short time. Through the comparison, it is possible to identify which products are on the shelf, which products are out of stock, and which products have changed in location or quantity.

[0065] Next, based on the matching results, the system will update the product information list in real time. For products on the shelf with correct information, the system will confirm their status and maintain the original record; for out-of-stock products, the system will mark them as out-of-stock in the list and may trigger a replenishment reminder; for products whose location or quantity has changed, the system will update their location and quantity information to ensure that the information in the database is consistent with the actual situation.

[0066] In addition, step S3 also involves the handling of abnormal situations. For example, if the system finds that the recognition result is significantly different from the information in the database, such as mismatched goods or abnormal quantities, the system may mark these abnormal information and notify management personnel for manual verification. This abnormal handling mechanism helps to ensure the accuracy and reliability of the system and avoid management confusion caused by misidentification or data errors.

[0067] It can be seen from the above technical solutions that a shelf commodity management method and related equipment provided by the embodiment of the present application first obtains a shelf image through a device-side interface, and obtains a shelf detection map according to a predefined recognition area configuration. This step solves the problem of commodity offset, ensures that the system can accurately obtain the commodity area to be identified, and improves the accuracy of recognition. After that, the shelf detection map is input into the trained commodity recognition model, and the commodity recognition model is used to perform a single forward propagation and non-maximum suppression process to generate a commodity recognition result. The commodity recognition model can efficiently identify the types and quantities of commodities on the shelf. Finally, the commodity recognition result is matched with the commodity information list in the application-side database, and the commodity information list is updated according to the matching result. This step realizes real-time monitoring and management of shelf commodities, and can promptly discover problems such as out-of-stock and placement errors. The present application has higher accuracy, reliability and management efficiency, can effectively solve the problems existing in the existing scheme, and provides a new solution for the shelf management field.

[0068] In some embodiments of the present application, the process of step S3, matching the commodity identification result with the commodity information list in the application-side database, and updating the commodity information list according to the matching result is introduced, which may specifically include:

[0069] Step S31, matching the commodity recognition result with the commodity information list in the application-side database in terms of shelf coordinate dimension, commodity area dimension and commodity category dimension respectively.

[0070] Specifically, the product recognition result is matched with the product information list in the application-side database in multiple dimensions, and these dimensions include but are not limited to shelf coordinate dimension, product area dimension, and product category dimension.

[0071] Shelf coordinate dimension: Compare the location information of the product in the recognition result with the location information recorded in the database to ensure that the coordinates of the two on the shelf are similar or consistent.

[0072] Product area dimension: Check whether the product is located in the area to which it belongs, such as a specific brand area, promotion area, etc., or a specific coordinate area range to ensure that the product is not mistakenly placed in a non-specified area.

[0073] Product category dimension: Verify whether the product category in the recognition result matches the record in the database, so as to facilitate the subsequent update of product information category and replenishment confirmation.

[0074] Step S32: If the matching degree of each dimension meets the preset deviation, the product information list is updated according to the product recognition result.

[0075] Specifically, if the matching degree of the product recognition results in each dimension meets the system preset deviation range (that is, the matching degree of each dimension meets the preset deviation), the product information list will be updated according to the recognition results. This includes updating the location, quantity, status and other information of the product to ensure that the information in the database is consistent with the actual situation.

[0076] Step S33: If there is a target matching dimension that does not meet the preset deviation, the target product information is analyzed based on the target matching dimension, and the target product corresponding to the target product information is determined as a misplaced product.

[0077] Specifically, if it is found that there are target matching dimensions that do not meet the preset deviations (that is, there are one or some dimensions whose matching degree is lower than the preset deviations), the target product information will be further analyzed based on these target matching dimensions. Specifically, the system will check the information of the target product and compare it with the records in the database to determine whether there is a misplacement. Once a product is confirmed to be a misplaced product (that is, its location, region or category does not match the records in the database), the system will mark it as a misplaced product and may trigger a series of subsequent operations, such as notifying management personnel to conduct verification and adjustments, and recording misplacement events for subsequent analysis. This mechanism helps to promptly detect and correct incorrect placement on shelves, ensuring the accuracy and standardization of product display.

[0078] On this basis, the product information list may also include the minimum quantity of each category;

[0079] The process of matching the commodity identification result with the commodity information list in the application-side database may also include:

[0080] Step S34: Match the quantity of each commodity in the commodity identification result with the minimum quantity of each category recorded in the commodity information list, confirm out-of-stock commodities whose quantity is lower than the minimum quantity of the corresponding category, and provide replenishment reminders for the out-of-stock commodities.

[0081] Specifically, setting a minimum quantity for each category is used to ensure that the goods on the shelves are always maintained at a certain inventory level to meet customer demand and avoid out-of-stock situations. At this time, in the process of matching the product identification results with the product information list in the application-side database, the quantity of each product in the product identification results will be compared one by one with the minimum quantity of each category recorded in the product information list. This comparison process involves checking the minimum inventory threshold set for each category of goods in the database to ensure that the actual inventory quantity is not lower than this threshold. If the system finds that the inventory quantity of a certain product is lower than the minimum quantity threshold of the category to which it belongs, the product will be automatically marked as out-of-stock, and the replenishment reminder function will be immediately triggered.

[0082] The inventory management capabilities of shelf merchandise management have been significantly improved by introducing category minimum quantity settings and out-of-stock merchandise identification and replenishment reminder functions. This optimization not only helps reduce the occurrence of out-of-stock situations, but also improves inventory turnover and operational efficiency.

[0083] In some embodiments of the present application, further analysis can be performed by statistically analyzing the identification results of each product within a predetermined analysis period, and the product promotion plan and product placement plan can be adjusted and designed to enhance customer experience, improve sales efficiency, and promote overall sales growth.

[0084] 1. Product promotion plan:

[0085] ① Count the identification results of each product within the predetermined analysis period;

[0086] ② generating a first adjusted commodity set according to the quantity fluctuation of each category of commodities in the commodity identification results within the predetermined analysis period and in combination with the commodity sales volume, wherein the first adjusted commodity set records each commodity whose quantity fluctuation frequency and sales volume ratio is greater than a first predetermined value;

[0087] ③ Design a product promotion plan based on the first adjusted product set.

[0088] Specifically, by analyzing the product recognition results within a predetermined analysis period, the aim is to identify products that are frequently viewed by customers but have low conversion rates (i.e., purchase rates), and to design promotional plans accordingly to increase sales. Specifically, the system will collect and count product recognition data on the shelves on a regular basis (e.g., weekly, monthly, etc.). Analyze the quantity fluctuations of each category of products based on the statistical results. Quantity fluctuations reflect the situation in which products are taken from the shelves by customers. A high frequency of fluctuations means that the product is frequently taken (frequently viewed).

[0089] Next, the ratio of the quantity fluctuation frequency to the sales volume is calculated. The larger the ratio, the lower the final purchase conversion rate, although the product is frequently taken. This type of product may be of interest to customers but failed to be purchased for some reason (such as price, packaging, substitutes, etc.). The system includes these products with a ratio greater than the first predetermined value into the first adjusted product set as the focus of subsequent promotion plan design.

[0090] For the products in the first adjusted product set, promotion plans can be intelligently designed based on historical sales data, customer preferences, market competition and other factors. Promotion plans may include price reduction promotions, gift combinations, limited-time discounts, member-exclusive discounts and other forms, aiming to stimulate customers' desire to buy and increase the conversion rate of these products. At the same time, the system can also customize personalized promotion information based on the characteristics of the product and the target customer group to accurately reach potential customers.

[0091] 2. Product placement plan:

[0092] ① Count the identification results of each product within the predetermined analysis period;

[0093] ② According to the quantity fluctuation of each category of commodities in the commodity identification results within the predetermined analysis period, combined with the commodity sales volume and commodity location information, a second adjusted commodity set is generated, wherein the second adjusted commodity set records each commodity whose quantity fluctuation frequency to sales volume ratio is less than a second predetermined value and whose commodity location belongs to a non-focus area;

[0094] ③ Design a merchandise placement plan based on the second adjusted merchandise set.

[0095] Specifically, by analyzing the commodity identification results within a predetermined analysis period, the commodity placement strategy is optimized, especially for those commodities with high sales conversion rates but located in non-focus areas. The larger the ratio of quantity fluctuation frequency to sales volume, it usually means that although the commodity is frequently paid attention to or taken, the proportion converted into actual sales is relatively low. On the contrary, the smaller the ratio, the higher the sales conversion rate, that is, the stronger the customer's willingness to buy the commodity. Since the focus is on those commodities that are located in non-focus areas but have high sales conversion rates (that is, customers tend to buy once they notice them), by screening out commodities with a smaller ratio of quantity fluctuation frequency to sales volume, that is, less than the second predetermined value, and whose location belongs to the non-focus area. Therefore, the second adjusted commodity set records commodities that can still achieve high sales with relatively less attention. Although these commodities are located in locations that are not easily noticed, they have a higher purchase conversion rate once they are discovered.

[0096] Design a more reasonable placement plan for the products in the second adjustment product set, aiming to improve the visibility of these products to customers, including moving them to the focal area of ​​the shelf, increasing the product display area, etc., to attract more customers' attention and promote sales.

[0097] A shelf commodity management device provided in an embodiment of the present application is described below. The shelf commodity management device described below and the shelf commodity management method described above can refer to each other.

[0098] See also Figure 3 , Figure 3 This is a schematic diagram of a shelf merchandise management device disclosed in an embodiment of the present application.

[0099] like Figure 3 As shown, the shelf commodity management device may include:

[0100] The image acquisition unit A1 is used to acquire the shelf image through the device end interface and obtain the shelf detection image according to the predefined recognition area configuration;

[0101] The commodity recognition unit A2 is used to input the shelf detection image into the trained commodity recognition model, and use the commodity recognition model to perform a single forward propagation and non-maximum suppression process to generate a commodity recognition result;

[0102] The matching and updating unit A3 is used to match the commodity identification result with the commodity information list in the application-side database, and update the commodity information list according to the matching result.

[0103] It can be seen from the above technical solutions that a shelf commodity management method and related equipment provided by the embodiment of the present application first obtains a shelf image through a device-side interface, and obtains a shelf detection map according to a predefined recognition area configuration. This step solves the problem of commodity offset, ensures that the system can accurately obtain the commodity area to be identified, and improves the accuracy of recognition. After that, the shelf detection map is input into the trained commodity recognition model, and the commodity recognition model is used to perform a single forward propagation and non-maximum suppression process to generate a commodity recognition result. The commodity recognition model can efficiently identify the types and quantities of commodities on the shelf. Finally, the commodity recognition result is matched with the commodity information list in the application-side database, and the commodity information list is updated according to the matching result. This step realizes real-time monitoring and management of shelf commodities, and can promptly discover problems such as out-of-stock and placement errors. The present application has higher accuracy, reliability and management efficiency, can effectively solve the problems existing in the existing scheme, and provides a new solution for the shelf management field.

[0104] Optionally, the commodity identification model includes a format conversion network, a forward propagation network, and a deduplication detection network;

[0105] The format conversion network is used to parse the input shelf detection image into binary image data, and generate an image matrix based on the binary image data conversion;

[0106] The forward propagation network performs commodity detection according to the image matrix through a single forward propagation to obtain a commodity detection result;

[0107] The deduplication detection network performs non-maximum suppression processing on the commodity detection result to generate a commodity recognition result.

[0108] Optionally, the matching update unit may include:

[0109] A dimension matching unit, used to match the commodity recognition result with the commodity information list in the application-side database in terms of shelf coordinate dimension, commodity area dimension and commodity category dimension respectively;

[0110] A list updating unit, configured to update the commodity information list according to the commodity identification result when the matching degree of each dimension meets the preset deviation;

[0111] The misplacement confirmation unit is used to analyze the target commodity information based on the target matching dimension when there is a target matching dimension that does not meet the preset deviation, and determine the target commodity corresponding to the target commodity information as a misplaced commodity.

[0112] Optionally, the product information list also includes a minimum quantity for each category;

[0113] The matching update unit may further include a replenishment confirmation unit, which is used to match the commodity quantity of each commodity in the commodity identification result with the minimum quantity of each category recorded in the commodity information list, confirm out-of-stock commodities whose quantity is lower than the corresponding minimum quantity of the category, and provide replenishment reminders for the out-of-stock commodities.

[0114] Optionally, the shelf commodity management device may further include:

[0115] A result statistics unit, used to count the identification results of each commodity within a predetermined analysis period;

[0116] A first set unit is used to generate a first adjusted commodity set according to the quantity fluctuation of each category of commodities in the commodity identification results within the predetermined analysis period and in combination with the commodity sales volume, wherein the first adjusted commodity set records each commodity whose quantity fluctuation frequency and sales volume ratio is greater than a first predetermined value;

[0117] A promotion design unit is used to design a product promotion plan based on the first adjusted product set.

[0118] Optionally, the shelf commodity management device may further include:

[0119] A result statistics unit, used to count the identification results of each commodity within a predetermined analysis period;

[0120] A second set unit is used to generate a second adjusted commodity set according to the quantity fluctuation of each category of commodities in the commodity identification results within the predetermined analysis period, in combination with commodity sales volume and commodity location information, wherein the second adjusted commodity set records each commodity whose quantity fluctuation frequency and sales volume ratio is less than a second predetermined value and whose commodity location belongs to a non-focus area;

[0121] A placement design unit is used to design a commodity placement plan based on the second adjusted commodity set.

[0122] The shelf commodity management device provided in the embodiment of the present application can be applied to shelf commodity management equipment. Figure 4 The hardware structure diagram of the shelf commodity management device is shown. Figure 4 ,The hardware structure of the shelf commodity management device may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;

[0123] In the embodiment of the present application, the number of the processor 1, the communication interface 2, the memory 3, and the communication bus 4 is at least one, and the processor 1, the communication interface 2, and the memory 3 communicate with each other through the communication bus 4;

[0124] The processor 1 may be a central processing unit CPU, or an application-specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention, etc.;

[0125] The memory 3 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), etc., such as at least one disk memory;

[0126] The memory stores a program, and the processor can call the program stored in the memory, wherein the program is used to:

[0127] Obtain shelf images through the device-side interface, and obtain shelf detection images based on pre-defined recognition area configurations;

[0128] Inputting the shelf detection image into the trained commodity recognition model, performing a single forward propagation and non-maximum suppression process using the commodity recognition model, and generating a commodity recognition result;

[0129] The commodity identification result is matched with the commodity information list in the application-side database, and the commodity information list is updated according to the matching result.

[0130] Optionally, the detailed functions and extended functions of the program may refer to the above description.

[0131] The embodiment of the present application further provides a readable storage medium, which may store a program suitable for execution by a processor, wherein the program is used to:

[0132] Obtain shelf images through the device-side interface, and obtain shelf detection images based on pre-defined recognition area configurations;

[0133] Inputting the shelf detection image into the trained commodity recognition model, performing a single forward propagation and non-maximum suppression process using the commodity recognition model, and generating a commodity recognition result;

[0134] The commodity identification result is matched with the commodity information list in the application-side database, and the commodity information list is updated according to the matching result.

[0135] Optionally, the detailed functions and extended functions of the program may refer to the above description.

[0136] The present application also provides a computer program product, including a computer program. When the computer program is executed by a processor, the execution method is as follows:

[0137] Obtain shelf images through the device-side interface, and obtain shelf detection images based on pre-defined recognition area configurations;

[0138] Inputting the shelf detection image into the trained commodity recognition model, performing a single forward propagation and non-maximum suppression process using the commodity recognition model, and generating a commodity recognition result;

[0139] The commodity identification result is matched with the commodity information list in the application-side database, and the commodity information list is updated according to the matching result.

[0140] Optionally, the detailed functions and extended functions of the program may refer to the above description.

[0141] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0142] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0143] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A shelf commodity management method, characterized in that: include: Obtain shelf images through the device-side interface, and obtain shelf detection images based on pre-defined recognition area configurations; Inputting the shelf detection image into the trained commodity recognition model, performing a single forward propagation and non-maximum suppression process using the commodity recognition model, and generating a commodity recognition result; The commodity identification result is matched with the commodity information list in the application-side database, and the commodity information list is updated according to the matching result.

2. The method according to claim 1, characterized in that: The commodity identification model includes a format conversion network, a forward propagation network, and a deduplication detection network; The format conversion network is used to parse the input shelf detection image into binary image data, and generate an image matrix based on the binary image data conversion; The forward propagation network performs commodity detection according to the image matrix through a single forward propagation to obtain a commodity detection result; The deduplication detection network performs non-maximum suppression processing on the commodity detection result to generate a commodity recognition result.

3. The method according to claim 1, characterized in that Matching the commodity identification result with the commodity information list in the application-side database, and updating the commodity information list according to the matching result, including: Match the product recognition result with the product information list in the application-side database in terms of shelf coordinate dimension, product area dimension, and product category dimension respectively; If the matching degree of each dimension meets the preset deviation, the product information list is updated according to the product recognition result; If there is a target matching dimension that does not meet the preset deviation, the target product information is analyzed based on the target matching dimension, and the target product corresponding to the target product information is determined as a misplaced product.

4. The method according to claim 3, characterized in that The product information list also includes the minimum quantity for each category; Matching the commodity identification result with the commodity information list in the application-side database also includes: The commodity quantity of each commodity in the commodity identification result is matched with the minimum quantity of each category recorded in the commodity information list, out-of-stock commodities whose quantity is lower than the minimum quantity of the corresponding category are confirmed, and replenishment reminders are issued for the out-of-stock commodities.

5. The method according to claim 1, characterized in that Also includes: Collect statistics on the identification results of each product within the predetermined analysis period; generating a first adjusted commodity set according to the quantity fluctuation of each category of commodities in the commodity identification results within the predetermined analysis period and in combination with the commodity sales volume, wherein the first adjusted commodity set records each commodity whose quantity fluctuation frequency and sales volume ratio is greater than a first predetermined value; A product promotion plan is designed based on the first adjusted product set.

6. The method according to claim 1, characterized in that Also includes: Collect statistics on the identification results of each product within the predetermined analysis period; Generate a second adjusted product set according to the quantity fluctuation of each category of products in the product identification results within the predetermined analysis period, combined with product sales volume and product location information, wherein the second adjusted product set records each product whose quantity fluctuation frequency to sales volume ratio is less than a second predetermined value and whose product location belongs to a non-focus area; A merchandise placement plan is designed based on the second adjusted merchandise set.

7. A shelf merchandise management device, characterized in that: include: An image acquisition unit, used to acquire a shelf image through a device-side interface and obtain a shelf detection image according to a predefined recognition area configuration; A commodity recognition unit, used for inputting the shelf detection image into a trained commodity recognition model, performing a single forward propagation and non-maximum suppression process using the commodity recognition model, and generating a commodity recognition result; The matching and updating unit is used to match the commodity identification result with the commodity information list in the application-side database, and update the commodity information list according to the matching result.

8. A shelf commodity management device, characterized in that: including memory and processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the shelf commodity management method according to any one of claims 1 to 6.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the shelf commodity management method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, each step of the shelf commodity management method according to any one of claims 1 to 6 is executed.

Images