An electronic price tag position recognition method and system

Abstract

The present invention relates to the field of electronic price tag management, and particularly to a method and system for identifying the position of an electronic price tag. The method includes the following steps: obtaining real-time image data from a retail environment, and enhancing the image data through a multi-modal sensor to obtain a standardized image sequence; denoising, enhancing, and optimizing the quality of the image sequence, and extracting image features through a deep convolutional neural network to generate candidate regions for the position of the electronic price tag; calculating the true position coordinates of the electronic price tag in combination with the image spatial information, and synchronously updating the commodity price; sending the updated electronic price tag position and price information to a cloud computing platform for storage, analysis, and management; based on real-time feedback and historical data, optimizing the image processing and position recognition strategies through reinforcement learning to generate adaptive adjustment parameters. The present invention can efficiently and accurately identify the position of the electronic price tag in a complex retail environment, and achieve real-time price synchronization update, with high automation and accuracy.

Description

Technical Field

[0001] The present invention relates to the field of new generation information technology and retail, and in particular to a method and system for identifying the position of an electronic price tag. Background Art

[0002] With the development of intelligence and automation in the retail industry, electronic price tags have become an important identification tool in modern retail environments. Electronic price tags can not only display product prices, promotional information, etc., but also interact with product management systems in real time to achieve automatic update and synchronization of price information. Especially in large-scale retail environments, efficient management and accurate identification of electronic price tags have become the key to improving operational efficiency and reducing labor costs. However, existing electronic price tag systems face technical challenges of accurate identification and real-time synchronization in complex and dynamic retail environments, especially under conditions such as lighting changes, product occlusion, and different viewing angles.

[0003] Current electronic price tag recognition methods mainly rely on image processing and computer vision technology, but due to the layout of goods in the retail environment, the wide variety of goods, and the interference of environmental factors, it is difficult for existing technologies to achieve efficient and accurate electronic price tag location recognition in a changing environment. In addition, although existing technologies can achieve electronic price tag location recognition in some specific environments, their accuracy and real-time performance in complex environments are still low, especially in the application of multimodal data fusion and deep learning algorithms.

[0004] Therefore, how to accurately identify the location of electronic price tags and achieve real-time price synchronization updates through multimodal sensor data and deep learning algorithms in a complex and dynamic retail environment is a core problem that needs to be solved in current technology. Solving this problem can not only improve the automation level of electronic price tag management, but also significantly improve the accuracy and stability of the system in a changing environment. Summary of the invention

[0005] The present invention provides an electronic price tag location recognition method and system to solve the problem of how to achieve accurate location recognition and real-time price synchronization update of electronic price tags in a complex and dynamic retail environment through multimodal sensor data and deep learning algorithms, so as to improve the automation and accuracy of electronic price tag management.

[0006] In order to solve the above technical problems, the present invention provides a method for identifying the position of an electronic price tag, comprising:

[0007] Acquire real-time image data from a retail environment and enhance the image data through a multimodal sensor to obtain a standardized image sequence;

[0008] Denoise and enhance the standardized image sequence, optimize the image quality, and extract image features through a deep convolutional neural network to generate a sequence of candidate regions for the positions of electronic price tags;

[0009] Based on the sequence of candidate regions for the positions of electronic price tags and combined with the spatial information of the image, calculate the true position coordinates of the electronic price tags; synchronously update the prices of the commodities according to the positions and price information of the electronic price tags;

[0010] Send the updated position and price information of the electronic price tags to the cloud computing platform for storage, analysis and management;

[0011] Based on real-time data feedback and historical environmental data, optimize the image processing and position recognition strategies through reinforcement learning, generate adaptive adjustment parameters, and track the subsequent recognition and update processes.

[0012] Further, before the step of acquiring real-time image data from the retail environment, it further includes:

[0013] Configure multiple sensors for acquiring image data under different lighting, perspectives and environmental changes, and the sensors include RGB image sensors, depth sensors and infrared sensors.

[0014] Further, the step of denoising and enhancing the image sequence specifically includes:

[0015] Use an image denoising algorithm to eliminate the noise in the image, and further perform image enhancement processing such as brightness and contrast.

[0016] Further, the step of extracting image features through a deep convolutional neural network specifically includes:

[0017] Input the image sequence into the deep convolutional neural network, extract local features in the image through the convolutional layer, and reduce the data dimension through the pooling layer to generate a feature map containing the key information of the image.

[0018] Further, the deep convolutional neural network includes multiple convolutional layers and pooling layers, extracts spatial features in the image layer by layer through the multi-layer network, and generates image features for position recognition. Further, the step of accurately calculating the position based on the sequence of candidate regions for the positions of the electronic price tags specifically includes:

[0019] Combined with the spatial information of the image, calculate the true position coordinates of the electronic price tags through perspective transformation and depth image information, and perform subsequent price synchronization updates according to the coordinate information.

[0020] Further, the step of calculating the position coordinates of the electronic price tag through perspective transformation and depth information further includes:

[0021] Use the internal and external parameters of the camera of the said image to correct the candidate area of the electronic price tag, so as to accurately calculate its three-dimensional spatial position.

[0022] Further, the step of sending the updated electronic price tag position and price information to the cloud computing platform for storage, analysis and management specifically includes:

[0023] Transmit the electronic price tag position and price information to the cloud computing platform through the network. The cloud computing platform is used for real-time storage and management of data, and provides data analysis and query interfaces.

[0024] Further, the step of optimizing the image processing and position recognition strategy through reinforcement learning specifically includes:

[0025] Based on real-time feedback data and historical environment data, optimize the image processing and position recognition models through reinforcement learning algorithms, and generate adaptive adjustment parameters to improve recognition accuracy and system adaptability.

[0026] Further, an electronic price tag position recognition system includes:

[0027] A data acquisition module, which is used to obtain image data related to the retail environment from a variety of sensors. The sensors include RGB image sensors, depth sensors, infrared sensors, etc., and collect commodity images and environmental parameters in the retail environment in real time through these sensors;

[0028] A data processing module, which is used to denoise, enhance, and standardize the image data obtained from the data acquisition module, and transmit the processed image data to the subsequent module;

[0029] A feature extraction and selection module, which is used to extract key features that can effectively identify the position of the electronic price tag from the processed image data, and screen out the most discriminative features;

[0030] A position calculation and synchronous update module, which is used to calculate the real position coordinates of the electronic price tag based on the extracted features of the candidate area of the electronic price tag position, combined with the spatial information of the image, and synchronously update the price information of the commodity; A cloud computing platform and a data storage module, which are used to store and manage all generated electronic price tag data, and provide data analysis, query, and multi-terminal access functions;

[0031] A real-time feedback and optimization module, which is used to receive the feedback data of each module, and dynamically adjust the image processing algorithm and recognition strategy according to the real-time data to optimize the system performance;

[0032] A user interaction and visualization module, which is used to provide an intuitive user interface, display the positions and price information of the electronic price tags, and support real-time data interaction and operation configuration;

[0033] A system monitoring and adaptive module, which is used to monitor the running status of the entire system in real time, and adjust the system parameters according to environmental changes through an adaptive learning mechanism to ensure the stability and adaptability of the system in different retail environments.

[0034] The key innovations of the present invention include:

[0035] (1) Multi-modal sensor data enhancement: By using multi-modal sensors such as RGB images, depth sensors, and infrared sensors, the system can obtain more comprehensive and reliable environmental data to ensure high-quality image data in complex retail environments.

[0036] (2) Deep convolutional neural network (CNN) feature extraction: Using a deep convolutional neural network to automatically extract key information in the image, effectively identify features such as the shape, size, and color of the electronic price tag, and optimize the accuracy of position recognition.

[0037] (3) Spatial information fusion and precise position calculation: Combining the spatial information of the image (such as perspective transformation, depth image, etc.), calculate the true position coordinates of the electronic price tag to improve the accuracy of position recognition.

[0038] The following are its main beneficial effects:

[0039] The method and system for identifying the position of an electronic price tag provided by the present invention can accurately identify the position of the electronic price tag and realize real-time synchronous update of price information in a complex and dynamically changing retail environment through multi-modal sensor data and deep learning algorithms. Compared with traditional image processing methods, the multi-modal data enhancement of the present invention and the precise position calculation combining the deep convolutional neural network (CNN) with spatial information effectively improve the accuracy of electronic price tag recognition. Especially in complex environments such as light changes, commodity occlusion, and different perspectives, it can ensure the efficiency and stability of position recognition. At the same time, through real-time feedback and optimization of historical data, the system can adaptively adjust the image processing and position recognition strategies, improving the long-term adaptability and stability of the system, and avoiding the problems of recognition failure and system performance degradation caused by environmental changes in traditional methods. In addition, synchronizing the updated electronic price tag data to the cloud computing platform for storage and management realizes efficient storage, analysis, and query of data, improving the overall operation and maintenance efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic flowchart of a method for identifying the position of an electronic price tag provided by an embodiment of the present application;

[0041] Figure 2 Block diagram of an electronic price tag position recognition system provided by an embodiment of the present application. Detailed implementation manners

[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs; the terms used in the description of the present application in this specification are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the description and claims of this application and the above drawings are intended to cover non-exclusive inclusion. The terms "first", "second", etc. in the description and claims of this application or the above drawings are used to distinguish different objects and not to describe a specific order.

[0043] Referring to "embodiment" herein means that a specific feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of this application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

[0044] Embodiment 1: Refer to Figure 1 , which is a schematic flowchart of an electronic price tag position recognition method and system provided by an embodiment of the present invention. This process may at least include steps S100 - S500:

[0045] S100. Obtain real-time image data from the retail environment and perform data enhancement through a multi-modal sensor to obtain a standardized image sequence.

[0046] S200. Denoise and enhance the image sequence, optimize the image quality, and extract image features through a deep neural network to generate a sequence of candidate regions for the electronic price tag position.

[0047] S300. Based on the sequence of candidate regions for the electronic price tag position, perform precise calculation of the electronic price tag position, determine the final position coordinates, and synchronously update the price information of relevant commodities.

[0048] S400. Send the updated electronic price tag position and price information to the cloud computing platform for storage, analysis, and management.

[0049] S500. Based on real-time data feedback and historical environmental data, optimize the image processing and position recognition strategies through reinforcement learning, generate adaptive adjustment parameters, and track subsequent recognition and update processes.

[0050] Step S100 includes at least steps S110 - S130:

[0051] S110: Obtain real - time image data from multiple sensors in a retail environment, covering different perspectives, lighting, and occlusion situations.

[0052] In a retail environment, multiple sensors are used to collect real - time image data. Specifically, the sensors include, but are not limited to, RGB image sensors, depth sensors, infrared sensors, temperature sensors, etc. Each sensor is responsible for obtaining different information dimensions:

[0053] RGB image sensor: Collect visual information in the retail environment, including the color, shape, size, etc. of goods;

[0054] Depth sensor: Provide depth information of objects, which can help judge object distance, stereo effect, and three - dimensional position;

[0055] Infrared sensor: Used to obtain the heat distribution in the retail environment, which helps to identify darker objects caused by insufficient light or reflection;

[0056] Temperature sensor: In specific situations, supplement information related to the ambient temperature, which helps in understanding image data in complex environments.

[0057] The sensors collect data in real - time in a synchronous manner, and the obtained image data covers different perspectives, lighting, and occlusion situations. These multi - dimensional data ensure that all information in the retail environment can be obtained for subsequent image fusion and analysis. The output of this step is a set of image data, denoted as where contains multi - modal data collected from different sensors.

[0058] S120: Enhance the obtained image data using multi - modal data.

[0059] After obtaining the multi - modal image data it is processed through multi - modal data enhancement techniques. Specifically, the image enhancement includes the following aspects:

[0060] For the images collected by the RGB image sensor, color balance and enhancement algorithms are adopted. For example, by adjusting brightness, contrast, saturation, etc., to ensure that the visual effect of the image is more prominent.

[0061] Based on the depth information provided by the depth sensor, perform depth enhancement on the RGB image, combine the three - dimensional information of the object, solve problems caused by occlusion or the spatial distribution of objects, and enhance the depth sense and details of the image.

[0062] For the image missing caused by insufficient light or reflection, utilize the information of the infrared sensor for enhancement. Ensure that in low-light or backlight environments, the image can still effectively represent the object features in the retail environment.

[0063] Through the enhancement of these multimodal data, generate an enhanced image data set, denoted as This data set compared with the original data Is improved in terms of visual clarity, depth information, and environmental adaptability. Specifically, the enhancement effect of the image data can be described by the following formula:

[0064]

[0065] Where, Is the enhancement function, representing the comprehensive processing of image enhancement; Is the original image data; Is the depth information; Is the infrared information.

[0066] S130: Standardize the enhanced image data using multimodal sensors.

[0067] After completing the data enhancement, next, perform standardization on the enhanced image data To ensure the stability and consistency of subsequent processing. The standardization process includes the following steps:

[0068] Adjust the image to a unified size (e.g., 224×224 pixels) to meet the input requirements of the deep learning model. This step ensures that all image data has the same spatial dimension for subsequent image feature extraction.

[0069] Unify the color space of the image, convert the image from the RGB color space to a standard color space (e.g., Lab or grayscale), thereby removing the interference of color on image recognition and ensuring consistency during feature extraction.

[0070] Normalize the pixel values to a certain range (e.g., between 0 and 1) to reduce the impact of brightness differences between different images and reduce the deviation caused by numerical inconsistencies.

[0071] Through these standardization operations, finally obtain the standardized image data set, denoted as Which can be directly used for subsequent image feature extraction and location recognition processing.

[0072] The standardization formula is as follows:

[0073]

[0074] Where, is the enhanced image data; μ is the mean of the image data; σ is the standard deviation of the image data.

[0075] This standardization operation ensures that the numerical distribution of the image meets the processing requirements, facilitating the processing of subsequent algorithms.

[0076] Connection between the front and rear steps: In the S110 stage, the acquired real-time image data is jointly collected by multiple sensors, covering different perspectives, lighting conditions, and occlusion situations. This data set provides the original input data for the subsequent enhancement step (S120).

[0077] In the S120 stage, the image data is enhanced by a multi-modal sensor to generate the enhanced image data , thereby improving the quality and spatial information of the image. These enhanced images provide the basis for the subsequent standardization (S130).

[0078] In the S130 stage, the enhanced image data is standardized to prepare for subsequent image feature extraction (such as feature extraction of depth features in a neural network), ensuring that the image data can be efficiently processed in subsequent algorithms.

[0079] Step S200 includes at least steps S210 - S230:

[0080] S210: Remove the noise interference in the image through an image denoising algorithm to improve the clarity and effectiveness of the image.

[0081] In step S210, the image sequence (the image data after standardization in the previous step) needs to remove the noise interference in the image through a denoising algorithm, thereby improving the clarity and effectiveness of the image. Specifically, the denoising process can be completed through the following steps:

[0082] Assume that the noise in the image data is Gaussian noise i.e., the image pixel value can be expressed as:

[0083]

[0084] where is the pixel value of the real image, is the noise.

[0085] Use classical denoising algorithms (such as Gaussian filtering, mean filtering, median filtering, bilateral filtering, etc.) to process the image. The specific steps are as follows:

[0086]

[0087] Among them, represents removing noise using a specific filter, where σ is the standard deviation of the filter, controlling the intensity of denoising. The image obtained after denoising is .

[0088] The signal-to-noise ratio (SNR) is used to evaluate the denoising effect. The denoising effect is measured by calculating the SNR before and after denoising. The formula for SNR is:

[0089]

[0090] Among them, represents the variance of the image or noise.

[0091] Finally, the denoising operation outputs the denoised image data , providing clear data input for subsequent enhancement processing.

[0092] S220: Adopt image enhancement technology to improve the prominence of image details and important regions.

[0093] In step S220, for the denoised image , adopt image enhancement technology to improve the details and prominence of important regions of the image. The enhancement processing steps specifically include:

[0094] By adjusting the brightness and contrast of the image, the important regions in the image become more obvious, especially the electronic price tag and its surrounding areas. The process can be completed by adjusting the contrast and brightness of pixel values:

[0095]

[0096] Among them, α is the contrast adjustment factor, μ is the mean value of the image, and β is the brightness adjustment factor. This operation can effectively enhance the contrast of the image, making the details of the image clearer.

[0097] For important detail regions in the image (such as the electronic price tag), use a sharpening algorithm (such as the Laplacian operator or the Sobel operator) to enhance the edge information, making these regions more prominent. The specific calculation formula is:

[0098]

[0099] Among them, k is the sharpening factor, represents the second-order gradient of the image (or the Laplacian operator), used to enhance edge details.

[0100] For some regions of interest in the image (such as the region where the electronic price tag is located), local enhancement techniques can be applied, and methods such as regional weighted average or Gaussian pyramid can be used to enhance local details:

[0101]

[0102] Among them, is the weight coefficient, and N is the number of pixel points in the local region.

[0103] Finally, the enhanced image data provides the details and key information of the image, especially enhancing the prominence of the electronic price tag region.

[0104] S230: Input the enhanced image sequence into a deep convolutional neural network to extract the key features in the image.

[0105] In step S230, the enhanced image data is input into a deep convolutional neural network (DCNN) to extract the key features in the image for subsequent electronic price tag position recognition.

[0106] First, design a convolutional neural network structure suitable for image feature extraction. The network consists of multiple convolutional layers, pooling layers, activation functions, and fully connected layers. The convolutional layer is used to extract local features from the image, and the pooling layer is used for downsampling to reduce the image size and retain important features.

[0107] Input the enhanced image into the convolutional neural network for a series of convolution and pooling operations. The convolution operation of each layer can be expressed as:

[0108]

[0109] Among them, is the feature map of the i-th layer, is the convolution kernel, * represents the convolution operation, is the bias, and σ(·) is the activation function (such as ReLU).

[0110] After multiple convolution and pooling operations, the high-level features extracted by the network contain the key information in the image, such as the position features of the electronic price tag. The final output of feature extraction can be used for subsequent generation of candidate regions for the electronic price tag position.

[0111] Finally, the feature map obtained through the deep convolutional neural network provides a sequence of candidate regions for the electronic price tag position, preparing for subsequent precise position calculation and synchronous update.

[0112] Connection between the previous and subsequent steps: In step S210, the image data is denoised, and the denoised image is output , providing clearer data for subsequent image enhancement.

[0113] In step S220, the denoised image is further enhanced to generate the enhanced image , improving the prominence of details and key information.

[0114] In step S230, the enhanced image data is input into a deep convolutional neural network to extract the key features in the image , which are used to generate a sequence of candidate regions for the positions of electronic price tags.

[0115] Step S300 includes at least steps S310 - S330:

[0116] S310: Use a computational model to further analyze and process the candidate regions for the positions of electronic price tags, and combine the spatial information of the image to calculate the true position coordinates of the electronic price tags.

[0117] In step S310, the sequence of candidate regions for the positions of electronic price tags (candidate regions obtained from the image features extracted by the deep convolutional neural network previously) needs to be further analyzed and processed, and the true position coordinates of the electronic price tags are calculated by combining the spatial information of the image. The specific steps are as follows:

[0118] First, based on the candidate regions for the positions of electronic price tags obtained in the previous step , by analyzing the spatial information of the image (such as perspective transformation in the image, the position and angle of the camera, etc.), the two - dimensional coordinates of each candidate region in the image can be obtained. The two - dimensional coordinates correspond to the positions on the image plane.

[0119] By combining the image data obtained from multiple sensors , as well as the internal and external parameters of the camera of the image (such as focal length, optical center, rotation matrix, etc.), a spatial geometry model (such as perspective transformation, camera matrix calculation, etc.) can be used to perform spatial correction on the image, so as to obtain the true three - dimensional positions of each candidate region ri. The specific calculation formula is:

[0120] Among them, is the three - dimensional position coordinate of the electronic price tag, K is the internal parameter matrix of the camera, R and T are the rotation matrix and translation matrix of the camera respectively, is the candidate region The pixel coordinates (assuming normalization).

[0121] Using the known camera parameters and image information, the coordinate calculation can be further optimized. By combining multi-view information (e.g., using stereo vision or optical flow method), more accurate positions of the electronic price tags can be obtained. Through this spatial information, the true spatial coordinates of each candidate region of the electronic price tag can be accurately calculated.

[0122] Finally, the calculated true position coordinates of the electronic price tag will be used as the input for subsequent steps.

[0123] S320: Combine the calculated accurate positions of the electronic price tags, obtain the price information of the corresponding products, and synchronously update the price information with the positions of the electronic price tags.

[0124] In step S320, combine the accurate positions of the electronic price tags calculated in the previous step , obtain the corresponding product price information, and synchronously update the price information with the positions of the electronic price tags. The specific steps are as follows:

[0125] The price information can be obtained in various ways. For example, by means of barcode scanning, RFID technology, querying the product database, etc., obtain the product price information corresponding to each electronic price tag from the product management system. Assume that the price information of a certain product is , then for each position of the electronic price tag the corresponding price information can be found , where i represents the i-th electronic price tag.

[0126] Through the data synchronization mechanism, associate the calculated positions of the electronic price tags and the corresponding price information , and synchronously update them to the electronic price tag system. Specifically, the positions of the electronic price tags and the price information can be sent to the cloud platform or the management system through the network:

[0127]

[0128] The synchronization process can be carried out using a standardized data format for transmission and storage. For example, use the JSON or XML format for the transmission of data packets.

[0129] During the synchronous update process, if there are situations of mismatched price information or incorrect positions of the electronic price tags, anomaly detection algorithms can be used for processing to ensure the correctness of the position information and price information of each electronic price tag in the system.

[0130] Finally, the synchronized and updated electronic price tag position and price information provide support for subsequent data management and storage.

[0131] S330: Generate new electronic price tag data based on the updated electronic price tag position and price information, and provide data support for subsequent data storage, transmission, and management

[0132] In step S330, based on the updated electronic price tag position and price information, generate new electronic price tag data, and provide data support for subsequent data storage, transmission, and management. The specific steps are as follows:

[0133] Based on the synchronized and updated electronic price tag position and price information , generate new electronic price tag data , which includes the coordinate information of the electronic price tag and the corresponding price information. The data format can be a standardized data structure, for example:

[0134]

[0135] Among them, represents the timestamp of data generation, is the unique identifier of the electronic price tag.

[0136] Store the generated electronic price tag data in a database or cloud platform for subsequent query, management, and analysis. During the storage process, the data will be indexed and classified according to the position and price of the electronic price tag for efficient retrieval and update.

[0137] The generated electronic price tag data also needs to be transmitted to terminal devices (such as electronic price tag displays) through the network. Specifically, the updated electronic price tag data is sent to the corresponding electronic tag terminal through wireless communication (such as Wi-Fi, Bluetooth, or RFID).

[0138] The system can continuously optimize the data management process of electronic price tags based on historical data and real-time feedback. For example, use a cloud computing platform for batch updates, anomaly detection, data archiving, etc.

[0139] Finally, through the above steps, new electronic price tag data is generated and provides support for subsequent storage, transmission, and management.

[0140] Connection between the previous and subsequent steps: In step S310, the candidate area is further analyzed, and the precise electronic price tag position is calculated by combining the spatial information of the image , providing the necessary position information for subsequent synchronized updates and data storage.

[0141] In step S320, the precise location of the item to which the price information pertains is synchronously updated, providing the necessary input for subsequent electronic price tag data generation.

[0142] In step S330, the updated electronic price tag location and price information and are used to generate new electronic price tag data , and provide support for subsequent data storage, transmission, and management.

[0143] Step S400 includes at least steps S410 - S430:

[0144] S410: Send the precisely calculated and updated electronic price tag location and price information to the cloud computing platform via the network.

[0145] In step S410, the precisely calculated and updated electronic price tag location and price information need to be transmitted to the cloud computing platform via the network. The specific process is as follows:

[0146] The electronic price tag location and price information will be packaged into data packets and transmitted in a standardized format. For example, the data can be encapsulated in JSON or XML format, and each data packet contains the electronic price tag location, price information, and other metadata (such as timestamp, device ID, etc.). Assume the data packet has the following format:

[0147]

[0148] where, represents the timestamp of data generation, is the unique identifier of the electronic price tag.

[0149] The formatted data packet is sent to the cloud computing platform via a wireless network (such as Wi-Fi, Bluetooth, or a dedicated communication protocol). During the transmission process, an encryption protocol (such as HTTPS) can be used to ensure data security. Use a reliable data transmission protocol (for example, TCP / IP protocol) to ensure the integrity and order of the data packets. During the transmission process, if a network failure or packet loss occurs, the system will automatically retry to ensure that the data arrives at the cloud platform accurately and error - free.

[0150] At the receiving end of the cloud computing platform, the data packet

[0151] ​It will be decoded and prepared for further processing. The data received by the cloud will be cached waiting for storage and analysis.

[0152] Through these steps, the updated electronic price tag location and price information are successfully transmitted to the cloud computing platform.

[0153] S420: Store the electronic price tag location and price information in the cloud computing platform

[0154] In step S420, the electronic price tag data sent to the cloud computing platform through the network needs to be stored in the cloud for subsequent query, analysis, and management. The specific steps are as follows:

[0155] In the cloud computing platform, the electronic price tag data will be stored in a distributed database or a cloud storage system. For efficient management, the data will be indexed according to the unique identifier of the electronic price tag and location For example, the electronic price tag data can be stored in a database table named TagData, and the table structure is:

[0156]

[0157] Within the cloud computing platform, the electronic price tag data can be classified and hierarchically stored according to different product categories, product locations, or timestamps. For example, different electronic price tag data can be stored according to regions (such as "Region A", "Region B") or product types (such as "Food", "Home Appliances") to improve the retrieval efficiency during subsequent data queries.

[0158] To ensure data security, the electronic price tag data stored on the cloud platform will be redundantly backed up. This can be achieved through multi-copy storage technology, backing up between different cloud servers or data centers to cope with hardware failures or network problems.

[0159] In addition to the location information and price information of the electronic price tag, the data packet also includes other metadata (such as timestamps, device IDs, etc.). These metadata will be used for subsequent data synchronization, query, and update.

[0160] Through these steps, the electronic price tag data D_{tagi} is stored in the cloud computing platform, ensuring the security and accessibility of the data in the cloud.

[0161] S430: Analyze and manage the stored electronic price tag data on the cloud computing platform, classify and synchronize the data, and process abnormal data

[0162] In step S430, the stored electronic price tag data will be analyzed and managed in the cloud computing platform. The specific steps are as follows:

[0163] By classifying and indexing the stored data, the electronic price tag data can be classified and managed according to the product type, location, or timestamp. For example, the data can be indexed by product type or region to facilitate the quick retrieval of electronic price tag data for specific categories or regions. The structure of the index table can be:

[0164] The electronic price tag data will be synchronously updated to the main database of the cloud computing platform. If a certain electronic price tag data changes (for example, price update or location change), the relevant data synchronization mechanism will synchronize the updated data to all copies and terminal devices in a timely manner to ensure the consistency of the cloud and terminal data. During the synchronous update process, a scheduled task or real-time update mechanism is used to ensure the timeliness of the data.

[0165] During the analysis process, the cloud platform will perform anomaly detection on the electronic price tag data. If data anomalies are detected (for example, missing price information or incorrect location data), the system will automatically trigger an anomaly handling mechanism. The abnormal data can be processed in the following ways:

[0166] Automatic correction: By docking with the product management system interface, automatically obtain the latest price information or location data, and update the abnormal data.

[0167] Manual review: For data that cannot be automatically corrected, the system will mark it as "abnormal" and notify the administrator for manual review and processing.

[0168] Data recording: The processing process of abnormal data will be recorded and a log file will be generated for subsequent analysis and optimization.

[0169] The electronic price tag data stored in the cloud platform can also be used for big data analysis to provide business decision support. For example, by analyzing historical price data, price fluctuation patterns can be discovered, and then the pricing strategy can be optimized. At the same time, based on the location information of the electronic price tag and product category data, analyze aspects such as the sales area and customer behavior.

[0170] Through the above steps, the stored electronic price tag data is effectively managed and analyzed to ensure the efficient operation of the system and the accuracy of the data.

[0171] Connection between the previous and subsequent steps: In step S410, the accurately calculated and updated electronic price tag location and price information It is transmitted to the cloud computing platform through the network, providing input for subsequent data storage and management.

[0172] In step S420, the electronic price tag data is stored in the cloud computing platform and efficiently managed through an index structure, providing a basis for subsequent analysis and synchronous update.

[0173] In step S430, the stored electronic price tag data will be classified, synchronously updated, and the accuracy and consistency of the data will be ensured through an abnormal data processing mechanism, providing support for subsequent query and analysis.

[0174] Step S500 at least includes steps S510 - S530:

[0175] S510: Evaluate the current image processing and position recognition strategies based on real - time feedback data, provide a basis for optimization, identify potential problems and make adjustments.

[0176] In step S510, the system first evaluates the effectiveness of the current image processing and position recognition strategies according to the real - time feedback data. The specific process is as follows:

[0177] The real - time feedback data includes the latest image data obtained through multi - modal sensors real - time electronic price tag position candidate regions , as well as the operating status and recognition accuracy data of the system (such as recognition accuracy error and processing time). This data can include real - time image sequences, real - time position coordinates, the differences between recognition results and true annotations, etc.

[0178] By analyzing the real - time feedback data and , evaluate the effectiveness of the current image processing and position recognition strategies. For example, calculate the effectiveness evaluation indicators of the image processing module (such as image denoising accuracy, enhancement effect, etc.), and the accuracy error of position recognition , the formula is as follows:

[0179]

[0180] where, is the true electronic price tag position, is the electronic price tag position predicted by the system, and n is the number of feedback data points.

[0181] According to the evaluation results, identify potential problems that may exist in the current strategy. For example, if the recognition accuracy error in the real - time feedback data If it exceeds the set threshold, there may be deficiencies in the image processing algorithm or the position recognition algorithm. At this time, the system will identify the key links that need to be adjusted, such as the image denoising algorithm, enhancement algorithm, or the network structure of the deep learning model.

[0182] According to the problem recognition results, corresponding adjustment suggestions are generated. These suggestions may include optimizing image denoising parameters, modifying image enhancement algorithms, adjusting hyperparameters of convolutional neural networks, etc., further providing a basis for subsequent model optimization.

[0183] Through the above steps, the system makes a preliminary evaluation of the image processing and position recognition strategies based on real-time feedback data and provides a basis for subsequent optimization.

[0184] S520: Optimize by combining historical environmental data, evaluate past experiences, adjust the parameters and recognition strategies of the model, and improve the long-term adaptability of the system.

[0185] In step S520, the system combines historical environmental data to optimize the model by evaluating past experiences and improve the long-term adaptability. The specific steps are as follows:

[0186] The historical environmental data includes historical image data , historical electronic price tag position data , and factors such as the noise level and light changes in the historical environment. By integrating historical data, information on the system's performance under different environmental conditions can be obtained, helping to evaluate the adaptability of the current strategy in different situations.

[0187] Conduct a retrospective evaluation of the system's performance in different environments through historical data. For example, historical image data and historical position data are used to evaluate the past recognition accuracy and image processing effects. Similar to step S510, calculate the recognition accuracy error in the historical environment :

[0188]

[0189] Among them, and are the true position and predicted position in the historical environment respectively, is the number of historical data points.

[0190] According to the recognition accuracy error in the historical data and other environmental factors (such as light changes, scene complexity, etc.), the system will adjust the parameters and strategies of the model. For example, if the historical data shows that the recognition accuracy is low in low-light environments, it may be necessary to adjust the image enhancement algorithm and increase processing means such as brightness enhancement.

[0191] Based on the feedback of historical environmental data, the system will propose specific optimization measures, such as adjusting the parameters of the image enhancement algorithm, modifying certain layers in the convolutional neural network, or increasing the adaptability to specific environments (such as optimizing the recognition of low-light environments by training the model). These adjustments will affect subsequent real-time strategies to improve the long-term stability and adaptability of the system.

[0192] Through the above steps and combined with historical environmental data, the system can optimize image processing and position recognition strategies under different environmental conditions and provide feedback for subsequent reinforcement learning.

[0193] S530: Utilize the reinforcement learning algorithm to optimize image processing and position recognition strategies through continuous training and feedback, and generate adjustment parameters adapted to different retail environments

[0194] In step S530, the system continuously optimizes image processing and position recognition strategies using the reinforcement learning algorithm to generate adaptive adjustment parameters. The specific process is as follows:

[0195] In the reinforcement learning framework, the system based on real-time feedback data and historical data conducts environmental interactions and feeds back the effect of the current strategy through the reward function. Specifically, define the reward function R as a comprehensive index of image recognition accuracy and processing efficiency, for example:

[0196]

[0197] where are the recognition accuracy errors in real-time and historical environments respectively, T is the processing time, is the weight coefficient.

[0198] In the reinforcement learning framework, the system continuously interacts with the environment (i.e., continuously executes image processing and position recognition), and optimizes the strategy based on the reward function R. For example, use Q-learning or Deep Q-Network (DQN) to update the policy function Q(S,a), where S is the state (such as current image and position data), a is the action (such as adjusting image enhancement parameters or the number of layers in the convolutional neural network), and the update formula is as follows:

[0199]

[0200] where is the learning rate, is the discount factor, S' is the new state, and a' is the next possible action.

[0201] Through continuous training, the adjustment parameters generated by the system are applied in image processing algorithms (such as denoising and enhancement) and location recognition strategies, thereby optimizing the overall system performance. For example, if the image enhancement algorithm fails to effectively improve the recognition accuracy in certain scenarios, reinforcement learning will guide the system to adjust the parameters and try new strategies until the best performance is achieved.

[0202] Through the above optimization process, the system continuously generates adjustment parameters adapted to different retail environments , and these parameters will dynamically adjust the system's image processing and location recognition strategies to adapt to different environmental conditions.

[0203] Through continuous training and feedback of the reinforcement learning algorithm, the system can gradually improve its adaptability and recognition accuracy in various environments.

[0204] Connection between the front and back steps: In step S510, the system evaluates the effect of the current strategy based on real-time feedback data, identifies potential problems, and provides a basis for optimization.

[0205] In step S520, combined with historical environmental data, past experience is evaluated, and the strategy and model parameters are adjusted to improve the long-term adaptability of the system.

[0206] In step S530, through the reinforcement learning algorithm, combined with real-time and historical data, the image processing and location recognition strategies are optimized to generate adjustment parameters adapted to different retail environments.

[0207] The key innovations of the present invention include:

[0208] (1) Multi-modal sensor data enhancement: By using multi-modal sensors such as RGB images, depth sensors, and infrared sensors, the system can obtain more comprehensive and reliable environmental data, ensuring high-quality image data in complex retail environments.

[0209] (2) Deep convolutional neural network (CNN) feature extraction: Using a deep convolutional neural network to automatically extract key information in the image, effectively identify features such as the shape, size, and color of the electronic price tag, and optimize the accuracy of location recognition.

[0210] (3) Spatial information fusion and precise location calculation: Combining the spatial information of the image (such as perspective transformation, depth image, etc.), calculate the true position coordinates of the electronic price tag to improve the accuracy of location recognition.

[0211] The following are its main beneficial effects:

[0212] The electronic price tag position recognition method and system provided by the present invention can accurately recognize the position of electronic price tags and achieve real-time synchronization and update of price information through multi-modal sensor data and deep learning algorithms in a complex and dynamically changing retail environment. Compared with traditional image processing methods, the precise position calculation combining multi-modal data enhancement and deep convolutional neural network (CNN) with spatial information in the present invention effectively improves the accuracy of electronic price tag recognition. Especially in complex environments such as light changes, product occlusion, and different viewing angles, it can ensure the efficiency and stability of position recognition. At the same time, through real-time feedback and optimization of historical data, the system can adaptively adjust image processing and position recognition strategies, improving the long-term adaptability and stability of the system, and avoiding problems such as recognition failures and system performance degradation caused by environmental changes in traditional methods. In addition, synchronizing the updated electronic price tag data to the cloud computing platform for storage and management realizes efficient storage, analysis, and query of data, improving the overall operation and maintenance efficiency of the system.

[0213] Embodiment 2: Figure 2 The structural block diagram of an electronic price tag position recognition system according to an embodiment of the present invention is shown. As Figure 2 shown, the system may include:

[0214] The data acquisition module 10 is used to obtain image data related to the retail environment from a variety of sensors. The sensors include RGB image sensors, depth sensors, infrared sensors, etc. Through these sensors, product images and environmental parameters (such as light, temperature, and humidity) in the retail environment are collected in real time. Specifically, the sensors can provide data under different viewing angles, lighting, and occlusion conditions to ensure the extensiveness and effectiveness of the collected data. In addition, the data acquisition module is also responsible for transmitting the collected data to the data processing module 20 in real time for subsequent processing.

[0215] The data processing module 20 is used to perform preliminary processing on the image data obtained from the data acquisition module 10. The specific processing steps include image denoising, enhancement, standardization, etc. to ensure the consistency and accuracy of the data. First, a denoising algorithm is used to eliminate noise and interference in the image, then the clarity of key information in the image is improved through enhancement processing, and finally the image data is standardized to ensure that the image data input to the subsequent module meets the predetermined format requirements. In addition, the data processing module is also responsible for transmitting the processed image data to the subsequent feature extraction and selection module 30.

[0216] The feature extraction and selection module 30 is responsible for extracting key features from the processed image data that can effectively identify the positions of the electronic price tags. Through deep learning algorithms such as deep convolutional neural network (CNN), local and global features in the image are automatically extracted, such as the shape, size, color, etc. of the electronic price tags. Further, using the feature selection algorithm, the most discriminative features are screened out to reduce redundant features and improve the accuracy and efficiency of subsequent position recognition. The extracted and selected features will be used as inputs and sent to the position calculation and synchronization update module 40.

[0217] Based on the feature data provided by the feature extraction and selection module 30, the position calculation and synchronization update module 40 performs precise calculation of the positions of the electronic price tags. By combining the spatial information of the image (such as perspective transformation, depth information, etc.) and the candidate regions of the electronic price tag positions obtained from the previous processing, the real position coordinates of the electronic price tags in the three-dimensional space are calculated, and the price information is synchronized and updated through the interface with the commodity management system. This module ensures the accuracy of the electronic price tag positions and at the same time guarantees the consistency between the prices of the commodities and their corresponding electronic price tag data.

[0218] The cloud computing platform and data storage module 50 is responsible for storing and managing all the electronic price tag data generated by the system. Through an efficient cloud storage solution, all the electronic price tag positions and price information (including real-time updated data) will be securely stored in the cloud, ensuring the persistence and reliability of the data. At the same time, the cloud computing platform can also analyze the stored data, provide query interfaces, and support data access and management on multiple terminals. Through data classification, synchronization update, and abnormal data processing mechanisms, the high availability and consistency of the electronic price tag data on the cloud platform are ensured.

[0219] The real-time feedback and optimization module 60 is responsible for receiving the feedback data from each module of the system and optimizing the position recognition strategy. By monitoring the running status and processing effect of the system in real time, this module can dynamically adjust the image processing algorithm and recognition strategy according to the actual running data (such as recognition accuracy, processing speed, etc.), thereby optimizing the system performance. This module can be fine-tuned based on the real-time feedback data to improve the stability and accuracy of the system in different environments and ensure the efficient recognition of the electronic price tags.

[0220] The user interaction and visualization module 70 provides an intuitive user interface for the system, supporting real-time data display and interactive operations. Users can view the recognition results, price information, and commodity status of the electronic price tags through this module and can adjust or configure the system according to their needs. Specifically, this module displays the positions and price information of the electronic price tags through a graphical interface to help users quickly understand and use the system. Through this module, users can effectively interact with the system, improving the decision-making efficiency and user experience.

[0221] The system monitoring and adaptive module 80 is responsible for real-time monitoring of the operating status of the entire system, including aspects such as hardware devices, data transmission, and algorithm execution. Through the adaptive learning mechanism, this module can automatically adjust various parameters of the system (such as image processing algorithms, model parameters, etc.) according to environmental changes and feedback data, ensuring the adaptability and stability of the system in different retail environments. This module can also trigger alarms and perform fault diagnosis in case of abnormal situations to ensure the continuous and efficient operation of the system.

[0222] Through the design of the above system structure modules, the electronic price tag position recognition method and system provided by the present invention have the following beneficial effects:

[0223] (1) High efficiency and accuracy: By combining multi-modal sensors and deep learning algorithms, the system can efficiently identify the positions of electronic price tags in complex retail environments and synchronously update the price information of goods. The system can process data under different lighting conditions, occlusions, and perspectives to ensure the accuracy of recognition.

[0224] (2) Real-time and flexibility: The system can collect and process image data in real time, and realize data storage and management through the cloud computing platform, ensuring the real-time update and high availability of price information. Through the real-time feedback and optimization mechanism, the system can adapt to different environmental changes and dynamically adjust strategies to improve the long-term adaptability of the system.

[0225] (3) Intelligence and self-adaptability: By combining reinforcement learning and historical environmental data, the system can continuously optimize image processing and position recognition strategies, generate self-adaptive adjustment parameters, and ensure that the system always maintains the best performance in the changing retail environment.

[0226] (4) User-friendliness: Through an intuitive user interface, users can easily view and manage electronic price tag information, perform operation configurations, improving the user experience and operation convenience.

[0227] Obviously, the embodiments described above are only a part of the embodiments of the present application, rather than all embodiments. The accompanying drawings show the preferred embodiments of the present application, but do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosed content of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements for some of the technical features. Any equivalent structure directly or indirectly using the content of the specification and drawings of the present application in other related technical fields is similarly within the scope of the patent protection of the present application.

Claims

1. An electronic price tag position recognition method, characterized in that Including the following steps: Obtain real-time image data from a retail environment through a multimodal sensor, where the multimodal sensor includes an RGB image sensor, a depth sensor, an infrared sensor, and a temperature sensor; Perform enhancement and normalization processing on the image data to generate a normalized image sequence, where the normalization processing includes resizing the image to a unified size, converting it to the Lab color space, and performing normalization processing. The expression for the normalization processing is: ； Among them, is the set of standardized image data; is the enhanced image data; μ is the mean of the image data; σ is the standard deviation of the image data; Perform denoising and enhancement processing on the normalized image sequence, extract image features through a deep convolutional neural network, and generate a sequence of candidate regions for the electronic price tag positions; Based on the sequence of candidate regions and the three-dimensional coordinates provided by the depth sensor, calculate the true three-dimensional position coordinates of the electronic price tag. The expression for the three-dimensional position coordinates is: ； Among them, is the three-dimensional position coordinate of the electronic price tag, K is the internal parameter matrix of the camera, and R and T are the rotation matrix and translation matrix of the camera respectively, is the pixel coordinate of the candidate region; Based on real-time data feedback and historical environment data, dynamically adjust the image processing parameters through a reinforcement learning algorithm, including the standard deviation of the filter and the contrast adjustment factor, to optimize the image processing and position recognition strategies. It includes the following steps: S510: Evaluate the current image processing and position recognition strategies according to the real-time feedback data, provide a basis for optimization, identify potential problems and make adjustments; S520: Combine historical environment data for optimization, evaluate past experience, adjust the parameters of the model and the recognition strategy, and improve the long-term adaptability of the system; S530: Use the reinforcement learning algorithm to optimize the image processing and position recognition strategies through continuous training and feedback, and generate adjustment parameters adapted to different retail environments.

2. The electronic price tag position recognition method according to claim 1, wherein The step of performing denoising and enhancement processing on the image sequence specifically includes: Use an image denoising algorithm to eliminate noise in the image, and further perform brightness and contrast image enhancement processing.

3. The method for identifying the position of an electronic price tag according to claim 1, wherein The step of extracting image features through a deep convolutional neural network specifically includes: Input the image sequence into the deep convolutional neural network, extract local features in the image through the convolutional layer, and reduce the data dimension through the pooling layer to generate a feature map containing the key information of the image.

4. The method for identifying the position of an electronic price tag according to claim 3, wherein The deep convolutional neural network includes multiple convolutional layers and pooling layers, extracts spatial features in the image layer by layer through the multi-layer network, and generates image features for position recognition.

5. The electronic price tag position recognition method according to claim 1, characterized in that, The step of performing precise position calculation based on the sequence of candidate regions for the electronic price tag positions specifically includes: Combine the spatial information of the image, calculate the true position coordinates of the electronic price tag through perspective transformation and depth image information, and perform subsequent price synchronization updates according to the coordinate information.

6. The method for identifying the position of an electronic price tag according to claim 5, wherein The step of calculating the position coordinates of the electronic price tag through perspective transformation and depth information further includes: Use the internal and external parameters of the camera of the image to correct the candidate region of the electronic price tag, so as to accurately calculate its three-dimensional spatial position.

7. The method for identifying the position of an electronic price tag according to claim 1, wherein The step of sending the updated electronic price tag position and price information to the cloud computing platform for storage, analysis, and management specifically includes: Transmit the electronic price tag position and price information to the cloud computing platform through the network. The cloud computing platform is used for real-time storage and management of data, and provides data analysis and query interfaces.

8. The method for identifying the position of an electronic price tag according to claim 1, wherein The step of optimizing the image processing and position recognition strategies through reinforcement learning specifically includes: Based on real-time feedback data and historical environmental data, optimize the image processing and position recognition models through reinforcement learning algorithms, and generate adaptive adjustment parameters to improve recognition accuracy and system adaptability.

9. An electronic price tag position recognition system, which is applied to the electronic price tag position recognition method according to any one of claims 1-8, is characterized in that, It includes: A data acquisition module, which is used to obtain image data related to the retail environment from multiple sensors. The sensors include RGB image sensors, depth sensors, and infrared sensors, and the commodity images and environmental parameters in the retail environment are collected in real time through these sensors; A data processing module, which is used to perform denoising, enhancement, and normalization processing on the image data obtained from the data acquisition module, and transmit the processed image data to the subsequent modules; A feature extraction and selection module, which is used to extract key features from the processed image data that can effectively identify the position of the electronic price tag, and screen out the most discriminative features; A position calculation and synchronous update module, which is used to calculate the real position coordinates of the electronic price tag based on the extracted candidate region features of the electronic price tag position, and combine the spatial information of the image, and synchronously update the price information of the commodity; A cloud computing platform and data storage module, which is used to store and manage all generated electronic price tag data, and provide data analysis, query, and multi-terminal access functions; A real-time feedback and optimization module, which is used to receive the feedback data of each module, and dynamically adjust the image processing algorithm and recognition strategy according to the real-time data to optimize the system performance; A user interaction and visualization module, which is used to provide an intuitive user interface, display the position and price information of the electronic price tag, and support real-time data interaction and operation configuration; A system monitoring and adaptive module, which is used to monitor the running status of the entire system in real time, and adjust the system parameters according to the environmental changes through an adaptive learning mechanism to ensure the stability and adaptability of the system in different retail environments.

Images