An image analysis and recognition system based on an unmanned vending cabinet

Abstract

The present invention provides an image analysis and recognition system based on an unmanned container, which includes a container image scanning unit, an image data analysis unit, a data association and matching unit, a difference parameter verification unit, and an image data verification unit; the container image scanning unit includes an inlet image scanning unit and an in-container image scanning unit. The inlet image scanning unit is arranged at the upper part and the left and right sides of the passage in the inlet of the container, and simultaneously scans the images of the items entering the container and sends the scanned data to the image data analysis unit; the in-container image scanning unit periodically scans the overall image of the shelves in the container and feeds the scanned data back to the image data analysis unit; the image data analysis unit has a commodity image database built in, and stores the image features of the commodities to be entered in the image database in advance, and matches the image data after receiving the image data uploaded by the inlet image scanning unit.

Description

Technical Field

[0001] The present invention belongs to the field of image recognition, and particularly relates to an image analysis and recognition system based on an unmanned vending cabinet. Background Art

[0002] With the rapid development of e-commerce and unmanned retail, the unmanned vending cabinet, as a new retail model, has gradually come into people's view. This model uses intelligent vending cabinet devices and image analysis and recognition technologies to achieve automatic identification and transaction of goods, providing a convenient shopping experience. However, the existing unmanned vending cabinet technologies based on image analysis still face many problems and limitations, restricting their wide application and development. First of all, the performance of existing image recognition technologies is often unsatisfactory in complex environments. Unmanned vending cabinets are usually under various lighting conditions, and the layout inside the cabinet and the way of placing goods may affect the accuracy of image recognition. When facing situations such as light changes, goods occlusion, or cluttered backgrounds, existing technologies are prone to recognition errors, resulting in the inability to accurately identify goods and affecting the user experience.

[0003] Traditional image recognition systems often rely on a large amount of labeled data for training, which may face the problem of insufficient data in practical applications. Especially for the image recognition of specific goods, obtaining high-quality labeled data is not only time-consuming and laborious, but may also lead to biases in model training, affecting the accuracy and reliability of recognition. In addition, existing systems usually lack the ability to quickly adapt to new goods. In the operation of unmanned vending cabinets, the frequency of goods updates is relatively high. Traditional image recognition systems need to retrain the model to adapt to new goods, resulting in long response times and low operation efficiency.

[0004] There are also hidden dangers in the security of existing unmanned vending cabinet systems. Due to the lack of effective identity verification mechanisms and anti-theft monitoring, it is easy for goods to be stolen or misappropriated, causing economic losses. Many existing systems do not consider the integration of user identity verification and anti-theft measures on the basis of image analysis, reducing the security and credibility of the system.

[0005] There are also deficiencies in the processing speed of the image analysis systems of existing unmanned vending cabinets. For application scenarios with high real-time requirements, traditional image processing algorithms may not be able to meet the needs of rapid recognition and response, thus affecting the user's shopping experience. Especially during peak periods, users' demands for rapid recognition and checkout of goods are more urgent, and the response speed of traditional systems may not meet the market's demands.

[0006] Many existing unmanned vending cabinet systems lack good user interaction designs. Users may encounter problems such as inconvenient operations and unclear information during use, resulting in a decline in the user experience. Such systems lacking humanized designs cannot effectively attract consumers' attention and usage.

[0007] Therefore, developing an image analysis and recognition system based on unmanned vending cabinets to overcome the deficiencies of the existing technology, improve recognition accuracy, response speed, and security has become an urgent need in the industry. The new system needs to be able to operate stably in complex environments, quickly adapt to the addition of new products, provide effective identity verification and anti-theft monitoring measures, and at the same time improve the user interaction design to enhance the shopping experience of consumers. By introducing advanced deep learning and computer vision technologies, combined with efficient algorithms and optimized system architectures, the future unmanned vending cabinet image analysis and recognition system is expected to achieve a higher level of intelligence and automation, promoting the development of the unmanned retail industry. Summary of the Invention

[0008] The present invention proposes an image analysis and recognition system based on unmanned vending cabinets. This image analysis and recognition system based on unmanned vending cabinets represents an important progress in the direction of more efficient, intelligent, and secure unmanned retail technology through its advanced design and functions, creating conditions for providing high-quality retail services, and is expected to significantly improve the operating efficiency of unmanned vending cabinets and the shopping experience of consumers.

[0009] The technical solution of the present invention is implemented as follows: An image analysis and recognition system based on unmanned vending cabinets includes a cabinet image scanning unit, an image data analysis unit, a data association and matching unit, a difference parameter verification unit, and an image data verification unit;

[0010] The cabinet image scanning unit includes an inlet image scanning unit and an in-cabinet image scanning unit. The inlet image scanning unit is arranged at the upper part and the left and right sides of the passage in the inlet of the vending cabinet, and simultaneously scans the images of the items entering the vending cabinet, and sends the scanned data to the image data analysis unit; The in-cabinet image scanning unit periodically scans the overall image of the shelves in the vending cabinet and feeds the scanned data back to the image data analysis unit;

[0011] The image data analysis unit is built-in with a commodity image database, and pre-stores the image features of the items to be entered in the image database. After receiving the image data uploaded by the inlet image scanning unit, it matches the image data. When the match is successful, the commodity enters the vending cabinet normally. When the match is unsuccessful, it synchronously sends the matching data and the image data to the data association and matching unit; When receiving the image uploaded by the in-cabinet image scanning unit, it identifies the empty spaces on the shelves, marks the empty spaces on the shelves, and outputs the information of the empty spaces;

[0012] The data association and matching unit calls the image features in the commodity database, and performs secondary analysis on the image features of each commodity. At least three feature regions are selected within the image features to establish a feature parameter deformation model. The image data that does not match in the incoming port image scanning unit and the commodity image database is repeatedly verified. The feature regions in the unmatched image data are compared with the feature regions of the corresponding commodities after secondary analysis in the commodity database, and a feature coincidence threshold is set for each type of commodity according to the commodity type. When the comparison data exceeds the set threshold, it is determined that the commodity comparison is successful, and the commodity is imported into the container; when the comparison data is lower than the set threshold, it is determined that the commodity comparison fails, a prohibited entry signal is sent to the unmanned container, and the commodity is imported into the waiting recycling shelf;

[0013] The difference parameter verification unit retrieves the feature parameter deformation model data in the association and matching unit, imports the difference parameters into the feature parameter deformation model, and performs timed verification on the feature regions in the feature parameter deformation model through the imported difference parameters. After each verification is completed, the verification result is synchronously fed back to the image data verification unit together with the verification time;

[0014] The image data verification unit classifies the verification results of the difference parameter verification unit, records the imported difference parameters and time data for the qualified verification results among them, records the change values of the feature parameters for the unqualified verification results among them, feeds back the change values to the user, and provides data reference for the set threshold.

[0015] Compared with the prior art, this image analysis and recognition system based on an unmanned container has multiple significant innovations and advantages. The system forms an efficient and complete commodity recognition and management system by constructing multiple functional modules, including a container image scanning unit, an image data analysis unit, a data association and matching unit, a difference parameter verification unit, and an image data verification unit. Traditional unmanned container systems often lack such a systematic design and usually rely on simple image capture and basic barcode scanning, unable to achieve comprehensive monitoring and management of commodities.

[0016] The setting of the container image scanning unit includes an incoming port image scanning unit and an in-container image scanning unit. This dual-scanning design greatly improves the monitoring accuracy of the system. The scanning at the incoming port ensures that each commodity entering the container can be recognized in real time, while the in-container scanning ensures continuous monitoring of the shelf status. Compared with the prior art, many systems can only perform a one-time detection after the commodity enters and cannot provide real-time updated information, resulting in inaccurate inventory management.

[0017] The image data analysis unit of the system has a built-in product image database, which can perform matching after receiving the image data scanned and uploaded at the inlet. This proactive matching mechanism improves the accuracy of product entry and avoids the situation of misreceiving unmatched products. Traditional systems generally rely on passive barcode scanning and cannot comprehensively analyze the appearance features of products, easily leading to misjudgment and incorrect operations. In addition, the data association and matching unit enhances the system's ability to recognize product features by performing secondary analysis on the product image features and establishing a deformation model of feature parameters. This in-depth analysis not only improves the matching accuracy but also enables the system to maintain high-efficiency recognition even when there are slight changes in the product appearance, which is lacking in existing technologies. Many traditional devices only rely on the original image data for simple matching and lack sensitivity to changes in image features, resulting in a decline in recognition rate in a dynamic environment.

[0018] The introduction of the differential parameter verification unit is a major highlight of the system. By regularly verifying the deformation model of feature parameters, the system can dynamically monitor changes in product features, ensuring the long-term reliability and accuracy of the system. Traditional systems often lack such a dynamic verification mechanism, resulting in a decline in recognition accuracy after long-term operation and affecting the overall performance. Fifth, the image data verification unit classifies the verification results, records the qualified and unqualified verification results, and provides an effective feedback mechanism. This function not only provides users with accurate data references but also enhances the transparency and trust of the system. Existing technologies are usually weak in this aspect and lack effective monitoring and feedback mechanisms, affecting the user experience.

[0019] The design of the system fully considers user needs and actual application scenarios. By providing detailed verification results and change values, the system not only helps users understand the product status in real time but also provides references for subsequent operations. This user-friendly design is often lacking in traditional unmanned vending cabinet systems, affecting user satisfaction and experience.

[0020] As a preferred implementation, the cabinet image scanning unit takes the center point of the channel inside the inlet of the cabinet as the focus, and installs the inlet image scanning unit on the upper part and the inner walls on the left and right sides of the channel inside the inlet of the cabinet around the focus. The inlet image scanning unit is a high-definition camera.

[0021] As a preferred implementation, after the inlet image scanning unit uploads the image data, the image data analysis unit preprocesses the received image. The preprocessing denoises, enhances, and standardizes the format of the image data, then extracts the color, shape, and texture information in the image data, compares it with the features in the built-in product image database, and calculates the similarity between the uploaded image and the product image features in the database through feature point matching.

[0022] As a preferred embodiment, before establishing the characteristic parameter deformation model, the geometric shape and color characteristics of the characteristic region are extracted, and model parameters are generated using an algorithm. The characteristic parameter deformation model is built based on the model parameters, and the characteristic coincidence threshold is set based on historical data and commodity characteristics analysis.

[0023] After adopting the above technical solutions, the beneficial effects of the present invention are as follows: The image analysis and recognition system based on the unmanned vending cabinet realizes efficient and accurate commodity recognition and management through the collaborative work of multiple modules. The vending cabinet image scanning unit can comprehensively scan the commodities entering the vending cabinet and the shelf situation inside the cabinet, ensuring real-time monitoring of the inventory status and timely updating of data. The image data analysis unit uses the built-in commodity image database for precise matching to ensure that only commodities meeting the standards can enter the vending cabinet smoothly, thereby reducing errors and losses. The data association and matching unit improves the flexibility and accuracy of commodity recognition through secondary parsing and the establishment of the characteristic parameter deformation model, and can effectively handle changes in the appearance of commodities. The difference parameter verification unit ensures the stability of the characteristic parameters through regular verification, further improving the reliability of the system. The image data verification unit classifies and records the verification results and promptly feeds them back to the user, enabling managers to better grasp the inventory changes and commodity characteristics and optimize the vending cabinet management process. Overall, the system not only improves the automation level of the unmanned vending cabinet but also enhances the accuracy and security of commodity management, providing strong support for the development of the unmanned retail industry. Description of the Drawings

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0025] Figure 1 It is a system flow block diagram of the present invention. Detailed Embodiments

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0027] Embodiment:

[0028] As Figure 1As shown in the figure, an image analysis and recognition system based on an unmanned container includes a container image scanning unit, an image data analysis unit, a data association and matching unit, a difference parameter verification unit, and an image data verification unit;

[0029] The container image scanning unit includes an inlet image scanning unit and an in-container image scanning unit. The inlet image scanning unit is arranged at the upper part and the left and right sides of the passage in the inlet of the container, and simultaneously scans the images of the items entering the container, and sends the scanned data to the image data analysis unit; the in-container image scanning unit periodically scans the overall image of the shelves in the container and feeds the scanned data back to the image data analysis unit;

[0030] The image data analysis unit has a commodity image database built in it, and stores the image features of the commodities to be entered in the image database in advance. After receiving the image data uploaded by the inlet image scanning unit, it matches the image data. When the matching is successful, the commodity enters the container normally. When the matching is unsuccessful, it synchronously sends the matching data and the image data to the data association and matching unit; when receiving the image uploaded by the in-container image scanning unit, it identifies the vacant positions on the shelves, marks the vacant positions on the shelves, and outputs the information of the vacant positions;

[0031] The data association and matching unit calls the image features in the commodity database, and performs secondary analysis on the image features of each commodity. At least three feature regions are selected in the image features to establish a feature parameter deformation model. It repeatedly verifies the image data that does not match between the inlet image scanning unit and the commodity image database, and compares the feature regions in the unmatched image data with the corresponding commodity feature regions after secondary analysis in the commodity database. And a feature coincidence threshold is set for each type of commodity according to the commodity type. When the comparison data exceeds the set threshold, it is determined that the commodity comparison is successful, and the commodity is imported into the container; when the comparison data is lower than the set threshold, it is determined that the commodity comparison fails, and a prohibited entry signal is sent to the unmanned container, and the commodity is imported into the shelf to be recycled;

[0032] The difference parameter verification unit retrieves the feature parameter deformation model data in the association and matching unit, imports the difference parameters into the feature parameter deformation model, and periodically verifies the feature regions in the feature parameter deformation model through the imported difference parameters. After each verification is completed, the verification result is synchronously fed back to the image data verification unit together with the verification time;

[0033] The image data verification unit classifies the verification results of the difference parameter verification unit, records the imported difference parameters and time data for the qualified verification results among them, records the change values of the feature parameters for the unqualified verification results among them, feeds back the change values to the user, and provides data reference for the set threshold.

[0034] The working principle of the image analysis and recognition system based on the unmanned container mainly revolves around links such as image scanning, data analysis, matching and verification, so as to realize intelligent commodity recognition and management. The container image scanning unit is divided into an inlet image scanning unit and an in-container image scanning unit. The inlet image scanning unit is responsible for scanning the commodities entering the container in real time and sending the obtained image data to the image data analysis unit. The in-container image scanning unit scans the shelves as a whole at regular intervals and feeds back the status and vacancy information of the commodities in the shelves. The image data analysis unit has a built-in commodity image database that stores the image features of various commodities. When receiving the image data uploaded by the inlet, the system will conduct a preliminary match. If the match is successful, the commodity will enter the container normally; if the match fails, the relevant data will be sent to the data association and matching unit for further processing. In the data association and matching unit, the system performs a secondary analysis on the unmatched commodity images, selects at least three feature regions to establish a feature parameter deformation model, and compares it with the feature regions in the commodity database. According to the different types of commodities, the system sets a feature coincidence threshold. If the comparison data exceeds the threshold, it is determined that the comparison is successful and the commodity enters the container; otherwise, it is determined that the comparison fails, a signal prohibiting entry is sent, and the commodity is imported into the shelf to be recycled.

[0035] The difference parameter verification unit is responsible for retrieving the feature parameter deformation model data, importing the difference parameters and performing regular verification, and feeding back the verification results and time to the image data verification unit. The image data verification unit classifies these verification results, records the qualified and unqualified results, feeds back the change value of the feature parameters to the user and provides data reference.

[0036] Compared with the current technology, the advantage of this working process lies in its high efficiency and accuracy. Through real-time image scanning and intelligent data analysis, the system can quickly identify commodities and conduct dynamic management, reduce manual intervention, and improve the commodity circulation efficiency. At the same time, the introduction of secondary analysis and the feature parameter deformation model enhances the system's adaptability to different environmental changes, thereby improving the recognition accuracy. The difference parameter verification mechanism ensures the continuous optimization of the system. By continuously recording and feeding back feature changes, it provides support for subsequent data analysis and forms a virtuous cycle. This comprehensive application of technology greatly improves the intelligent level of the unmanned container and provides a more efficient and reliable solution for the retail industry.

[0037] The container image scanning unit takes the center point of the channel inside the inlet of the container as the focus, and installs the inlet image scanning unit on the upper part and the inner walls on the left and right sides of the channel inside the inlet of the container around the focus. The inlet image scanning unit is a high-definition camera.

[0038] In the prior art, the image scanning systems of many unmanned vending cabinets usually adopt a single camera or a vision system with a fixed angle, resulting in a limited field of view for image capture when goods enter, which is likely to cause some goods to be blocked or inaccurately recognized. The image scanning unit at the inlet of this system is arranged around the center point of the passage inside the inlet of the vending cabinet, and multiple high-definition cameras are installed around the upper part and the inner walls on the left and right sides of the passage. This layout design with multiple angles significantly improves the coverage and clarity of image scanning. In this way, the system can capture multi-faceted images of the goods entering the vending cabinet in real time, avoiding the recognition blind spots that may be caused by a single perspective, thereby improving the accuracy and efficiency of goods recognition. In addition, the use of high-definition cameras ensures the image quality, can more accurately extract the features of the goods, and further enhances the effect of the recognition algorithm. By optimizing the layout of the scanning unit and improving the image quality, this system overcomes the limitations in the prior art and provides a more comprehensive and efficient goods recognition solution.

[0039] After the image scanning unit at the inlet uploads the image data, the image data analysis unit preprocesses the received images. The preprocessing denoises, enhances, and standardizes the format of the image data, then retrieves the color, shape, and texture information in the image data, and compares it with the features in the built-in goods image database, and calculates the similarity between the uploaded image and the goods image features in the database through feature point matching.

[0040] The technology has significant advantages compared with the prior art in the preprocessing process by the image data analysis unit after the image scanning unit at the inlet uploads the image data. In the prior art, there is usually a lack of systematicness and comprehensiveness in the image processing stage. Often, only basic denoising is performed, and the images cannot be effectively enhanced and format-standardized comprehensively, resulting in the subsequent feature extraction and matching processes being affected, and thus affecting the accuracy and efficiency of goods recognition. However, this technology ensures that the uploaded images have higher quality and consistency before feature extraction by integrating the preprocessing steps of denoising, enhancing, and format-standardizing, which lays a solid foundation for subsequent analysis. Specifically, through denoising, the interference information in the image can be effectively eliminated, and the clarity of the image can be improved; while the enhancement process helps to highlight the key features in the image, making the color, shape, and texture information more obvious, thereby improving the accuracy of feature extraction. In addition, format standardization ensures that images from different sources can be processed according to a unified standard, reducing the matching errors caused by format differences.

[0041] In the feature matching stage, the technology retrieves the color, shape, and texture information in the image data and compares it with the features in the built-in product image database. The systematicness of this process further improves the efficiency and accuracy of product recognition. Existing technologies often rely only on a single feature for matching and lack comprehensive analysis of multi-dimensional features, resulting in low recognition rates in complex environments. However, through the comprehensive use of multi-dimensional features, this technology can accurately capture the uniqueness of products and significantly improve the recognition accuracy even among similar products. Feature point matching calculates the similarity between the uploaded image and the product image features in the database, providing a scientific basis for subsequent automatic recognition and classification.

[0042] Through comprehensive preprocessing of image data and comprehensive analysis of multi-dimensional features, this technology is superior to existing technologies in terms of the accuracy, efficiency, and adaptability of product recognition, demonstrating more reliable and efficient characteristics in practical applications. The innovation of this method not only improves the success rate of product recognition but also provides more accurate data support for subsequent intelligent management and analysis, promoting the development and application of related technologies.

[0043] Before establishing the feature parameter deformation model, the geometric shape and color features of the feature region are extracted, and algorithm is used to generate model parameters. Based on the model parameters, the feature parameter deformation model is built, and the feature coincidence threshold is set based on historical data and product characteristics analysis.

[0044] The method for establishing the feature parameter deformation model has significant innovation and advantages compared with existing technologies. Traditional technologies are often limited to simple geometric shape or color feature extraction in the feature extraction stage and lack comprehensive analysis of the feature region, resulting in limitations in the adaptability and accuracy of the model in complex environments or cases of feature deformation. However, in this technology, before building the model, the geometric shape and color features of the feature region are comprehensively extracted first, ensuring a comprehensive and in-depth understanding of product characteristics. This comprehensive extraction method enables the model to more comprehensively reflect the actual features of products and lays a solid foundation for subsequent deformation model construction.

[0045] In the process of generating model parameters, the technology adopts advanced algorithms that can effectively convert the extracted features into model parameters. The intelligence and automation of this process significantly improve the efficiency and accuracy of model construction. In contrast, existing technologies often rely on manual experience or fixed rules for setting model parameters, which is prone to the influence of subjective factors and has low efficiency in dealing with complex data. However, the algorithm-driven approach of this technology can largely eliminate the errors caused by human intervention and make the generation of model parameters more scientific and objective.

[0046] The strategy of setting the feature overlap threshold based on the analysis of historical data and product characteristics is also a major highlight of this technology. In the prior art, a fixed threshold is usually used in feature matching, failing to consider the changes in product characteristics and historical data, which may lead to misidentification or missed identification in actual applications. In contrast, this technology dynamically adjusts the feature overlap threshold by analyzing historical data and combining the characteristics of the product, enabling the model to maintain good adaptability and accuracy in different environments and conditions. This flexible threshold setting method significantly improves the reliability of feature recognition, ensuring a high recognition rate for different products and under varying circumstances.

[0047] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An image analysis and recognition system based on an unmanned vending cabinet, characterized in that, It includes a container image scanning unit, an image data analysis unit, a data association and matching unit, a difference parameter verification unit, and an image data verification unit; The container image scanning unit includes an inlet image scanning unit and an in-container image scanning unit. The inlet image scanning unit is arranged in the upper part and the left and right sides of the channel in the inlet of the container, and simultaneously scans the images of the items entering the container, and sends the scanned data to the image data analysis unit; The in-container image scanning unit periodically scans the overall image of the shelves in the container and feeds the scanned data back to the image data analysis unit; The image data analysis unit has a commodity image database built in. The image features of the commodities to be entered are stored in the image database in advance. After receiving the image data uploaded by the inlet image scanning unit, it matches the image data. When the matching is successful, the commodity enters the container normally. When the matching is unsuccessful, the matching data and the image data are synchronously sent to the data association and matching unit; When receiving the image uploaded by the in-container image scanning unit, it identifies the vacant positions on the shelves, marks the vacant positions on the shelves, and outputs the information of the vacant positions; The data association and matching unit calls the image features in the commodity image database, and performs secondary analysis on the image features of each commodity. At least three feature regions are selected in the image features to establish a feature parameter deformation model. The image data that does not match between the inlet image scanning unit and the commodity image database is repeatedly verified. The feature regions in the unmatched image data are compared with the corresponding commodity feature regions after secondary analysis of the commodity image database, and a feature coincidence threshold is set for each type of commodity according to the commodity type. When the comparison data exceeds the set feature coincidence threshold, it is determined that the commodity comparison is successful, and the commodity is imported into the container; When the comparison data is lower than the set feature coincidence threshold, it is determined that the commodity comparison fails, a prohibited entry signal is sent to the unmanned container, and the commodity is imported into the shelf to be recycled; The difference parameter verification unit retrieves the feature parameter deformation model data in the association and matching unit, imports the difference parameters into the feature parameter deformation model, and periodically verifies the feature regions in the feature parameter deformation model through the imported difference parameters. After each verification is completed, the verification result is synchronously fed back to the image data verification unit together with the verification time; The image data verification unit classifies the verification results of the difference parameter verification unit, records the imported difference parameters and time data for the qualified verification results among them, records the change values of the feature parameters for the unqualified verification results among them, feeds back the change values to the user, and provides data reference for the set feature coincidence threshold.

2. The image analysis and recognition system based on an unmanned container according to claim 1, characterized in that: The inlet image scanning unit of the container image scanning unit is installed with the center point of the channel in the inlet of the container as the focus, surrounding the focus on the upper part and the inner walls of the left and right sides of the channel in the inlet of the container. The inlet image scanning unit is a high-definition camera.

3. The image analysis and recognition system based on an unmanned vending cabinet according to claim 1, wherein: After the image data is uploaded by the inlet image scanning unit, the image data analysis unit preprocesses the received image. The preprocessing denoises, enhances, and normalizes the format of the image data. Subsequently, the color, shape, and texture information in the image data is retrieved and compared with the features in the built-in commodity image database. The similarity between the uploaded image and the commodity image features in the database is calculated through feature point matching.

4. The image analysis and recognition system based on an unmanned container according to claim 1, wherein: Before establishing the feature parameter deformation model, the geometric shape and color features of the feature region are extracted, and the model parameters are generated using an algorithm. The feature parameter deformation model is built based on the model parameters, and the feature coincidence threshold is set based on the analysis of historical data and commodity characteristics.

Images