Intelligent article identification system and identification method

Abstract

The invention relates to the technical field of article identification, and discloses an intelligent article identification system and method, and the system comprises a data obtaining module which obtains the image data of a target article; the preprocessing module carries out denoising, normalization and contrast enhancement processing on the image data; the authentication area selection module automatically selects an authentication area; the feature extraction module identifies and extracts feature points in the authentication area, and generates corresponding high-dimensional vectors; the feature fingerprint processing module combines the high-dimensional vectors to generate feature fingerprints; the feature fingerprint matching module matches the generated feature fingerprints with feature fingerprints in a database and calculates the matching degree; and the output feedback module displays a matching result. According to the method, the flexibility and the reliability of article identification are enhanced, the dependence on traditional physical marks is reduced, the accuracy and the safety of article identification are improved, the flexibility and the adaptability are relatively high, the method can be widely applied to various occasions needing article identification, and a relatively high anti-counterfeiting capability is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of object identification, and in particular to an intelligent object identification system and an identification method. Background Art

[0002] In today's increasingly globalized world, the problem of counterfeit and substandard products has become particularly serious, posing a direct threat to businesses' economic interests and consumer safety. According to data from previous years, the global counterfeit market has reached approximately $4.2 trillion and is expected to continue growing. This continued growth not only undermines business profitability but also undermines consumer trust and loyalty to brands. The severity of this problem is reflected not only in economic losses but also in the damage it causes to social integrity and market order. The prevalence of counterfeit and substandard products has severely weakened fair competition in the market and has significantly impacted the operations of legitimate businesses.

[0003] While various anti-counterfeiting technologies, such as barcodes, etchings, seals, and methods using specialized materials, have been adopted to identify product origin, these methods still face numerous challenges. These markings are easily lost, damaged, or counterfeited, making it difficult to verify the product's performance, safety, and value. While traditional marking methods like barcodes and QR codes provide a certain level of product identification and verification, their physical properties make them susceptible to environmental influences such as wear, tear, and staining, which can lead to loss of their original identification function. Furthermore, tampering with or missing product labels often results in significant economic losses in both B2B and B2C transactions, reaching hundreds of billions of dollars annually. This figure is a conservative estimate; the actual situation is likely even more severe, as many losses are not accurately counted or reported. While these traditional anti-counterfeiting methods have served an important purpose in the past, the continuous advancement of counterfeiting technology has increasingly exposed their limitations, making them ineffective in combating the increasingly complex and diverse nature of modern counterfeit products.

[0004] Against this backdrop, object recognition technology plays a vital role in numerous fields, with applications spanning logistics tracking, retail management, security inspections, and production line automation. With the continuous advancement of technology and the increasing demands of society, object recognition technology has become an indispensable tool in many industries.

[0005] Traditional object identification methods typically rely on explicit identifiers such as barcodes and QR codes. While effective in many cases, these methods are susceptible to damage or forgery of the physical markings, significantly reducing the reliability and security of identification. These markings can wear, fade, or be tampered with over time, leading to reduced identification accuracy and, in turn, compromising the operational efficiency and security of the entire system. Summary of the Invention

[0006] The present invention provides an intelligent item identification system and an identification method. The high-resolution image of an item is captured by an image acquisition device, and feature extraction technology is used to analyze and extract the unique texture features on the surface of the item, which are then stored in a database for comparison during subsequent identification and verification processes. This enhances the flexibility and reliability of item identification, reduces the dependence on traditional physical markings, improves the accuracy and security of item identification, has high flexibility and adaptability, can be widely applied to various occasions where item identification is required, and provides higher anti-counterfeiting capabilities.

[0007] The present invention provides an intelligent item identification system, including a data acquisition module, a preprocessing module, an authentication area selection module, a feature extraction module, a feature fingerprint processing module, a feature fingerprint matching module, an output feedback module, and a database;

[0008] The data acquisition module is used to acquire the image data of the target item;

[0009] The preprocessing module is used to perform denoising, normalization, and contrast enhancement processing on the image data;

[0010] The authentication area selection module is used to automatically select the authentication area in the preprocessed image data to exclude background interference;

[0011] The feature extraction module is used to identify and extract feature points within the authentication area and generate corresponding high-dimensional vectors;

[0012] The feature fingerprint processing module is used to combine the high-dimensional vectors to generate a feature fingerprint;

[0013] The feature fingerprint matching module is used to match the generated feature fingerprint with the feature fingerprint in the database and calculate the matching degree;

[0014] The output feedback module is used to display the matching result; wherein, the matching result includes the matching degree, the number of matching pairs, and the detailed information of the matching points.

[0015] Furthermore, the data acquisition module includes a smart phone, an industrial camera, an X-ray machine, a near-infrared spectrometer, or other data capture devices.

[0016] Furthermore, in the preprocessing module, Gaussian filtering and median filtering methods are used for denoising to eliminate the noise in the image, and contrast stretching and histogram equalization methods are used for contrast enhancement to improve the contrast of the image data.

[0017] Furthermore, in the authentication area selection module,

[0018] Using Canny edge detection, Sobel operator and Laplacian operator, detect the edges and contours of the preprocessed image data to determine the boundary of the target item;

[0019] Using the GrabCut algorithm, through the user-specified initial bounding box, iteratively optimize to obtain the precise segmentation of the target item;

[0020] Using the graph cut algorithm, based on graph theory through the minimum cut / maximum flow algorithm, segment the image data into foreground and background;

[0021] Using the K-means clustering algorithm, cluster pixels into several categories by color, and select the clustering category containing the target item as the authentication area;

[0022] Using the live body cutting method, through energy minimization, iteratively optimize the contour to converge it to the boundary of the target item.

[0023] Furthermore, in the feature extraction module, identify and extract feature points within the authentication area using Harris corner detection, Shi-Tomasi corner detection, FAST, BRIEF, FREAK, HOG, GLOH, SIFT, SURF;

[0024] Generate the corresponding high-dimensional vector including:

[0025] Gradient calculation, by calculating the gradient direction and amplitude of the pixels around each feature point, obtain the local direction information;

[0026] Direction assignment, assign a direction and weight to each pixel according to the gradient direction and amplitude;

[0027] Histogram calculation, divide the local area into several sub-regions, statistically analyze the gradient directions within each sub-region to generate a direction histogram, and combine the direction histograms of all sub-regions into a high-dimensional vector, and the high-dimensional vector is the descriptor of the feature point.

[0028] Furthermore, the feature fingerprint processing module specifically includes:

[0029] Divide the image data into several image blocks, use the feature extraction method in the feature extraction module to extract several feature points and generate the corresponding high-dimensional vectors;

[0030] Combine the index of each image block and the descriptors of all feature points within the image block into a matrix, and write the index and descriptor matrix of each image block as a key-value pair; where the key is the index of the image block, and the value is the descriptor matrix of the feature points within the image block;

[0031] Combine the key-value pairs of all image patches into a dictionary as the feature fingerprint dictionary, store the feature fingerprint dictionary, and establish an indexing system to classify and store the feature fingerprint data; wherein, the storage format is.npz.

[0032] Further, the feature fingerprint matching module specifically includes:

[0033] Compare the generated feature fingerprint with the feature points of the feature fingerprint in the database by using the method of matching feature points; wherein, the method of matching feature points includes brute-force matching, K-nearest neighbor matching, bidirectional matching, and RANSAC.

[0034] Combine the methods of distance threshold, bidirectional verification, and block index alignment to screen out valid matching pairs according to the method of matching feature points, calculate the number of valid matching pairs, and obtain the overall matching metric value.

[0035] The present invention also provides an intelligent item identification method, based on the intelligent item identification system as described above, including the following steps:

[0036] Use an image acquisition device to acquire the original image data to be recognized and verified.

[0037] Preprocess the initial image data to obtain the initial image data; wherein, the preprocessing includes denoising, normalization, and enhancing contrast.

[0038] Automatically select the authentication area in the initial image data to exclude background interference.

[0039] Adopt Harris corner detection and FAST in the authentication area to identify and extract feature points, and generate corresponding high-dimensional descriptors.

[0040] Divide the initial image into several image patches, extract feature points in each image patch, generate a descriptor matrix, and form a key-value pair by combining the index of each image patch and the descriptor matrix, so as to form a dictionary with all key-value pairs to form a feature fingerprint.

[0041] Compare the feature fingerprint with the feature fingerprint in the database by using a matching algorithm to screen out valid matching pairs, and calculate the number of valid matching pairs to obtain the overall matching metric value as the matching degree.

[0042] If the matching degree is lower than the preset threshold, it is determined that the two images belong to the same item; otherwise, it is determined that they belong to different items, and finally output the matching result, including the matching degree, the number of matching pairs, and the detailed information of the matching points.

[0043] The present invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the above method are implemented.

[0044] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0045] The beneficial effects of the present invention are as follows:

[0046] First of all, the present invention has high efficiency and accuracy. Through advanced feature extraction and matching algorithms, the system can quickly and accurately identify and verify items, reducing the time and error rate of manual operations. Secondly, the present invention has high flexibility and adaptability, and can be customized and optimized according to different application scenarios and requirements, and is applicable to various complex working environments. In addition, the present invention also has good scalability. With the development of technology and the change of application requirements, the system can be continuously upgraded and expanded to meet future application needs. The modular design enables it to easily integrate new functions and technologies and adapt to the changing market demands. In short, through efficient and accurate identification and verification, not only the work efficiency and safety are improved, but also the cost is reduced and the user experience is enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic structural diagram of the intelligent item identification system in the present invention.

[0048] Figure 2 It is a schematic structural diagram of using the intelligent item identification system in combination with a remote database in the present invention.

[0049] Figure 3 It is a schematic diagram of the actual operation of the intelligent item identification system for photographing and registering a memory module in the present invention.

[0050] Figure 4 It is a schematic flowchart of the intelligent item identification method in the present invention.

[0051] Figure 5 It is a schematic flowchart of item matching in the intelligent item identification method in the present invention.

[0052] Figure 6 It is a schematic flowchart of the feature extraction step in the present invention.

[0053] Figure 7 It is a schematic flowchart of feature fingerprint generation and storage in the present invention.

[0054] Figure 8 It is a schematic internal structure diagram of a computer device according to an embodiment of the present invention.

[0055] The realization, functional features, and advantages of the present invention will be further described in conjunction with embodiments with reference to the accompanying drawings. Detailed implementation manners

[0056] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0057] The present invention is a system and method for quickly and accurately identifying and verifying items using image processing and feature extraction technologies.

[0058] To improve the accuracy and security of identification, the present invention proposes an authentication and identification method based on the unique features of the item surface. This method captures high-resolution images of the item through a high-definition camera and uses advanced feature extraction technologies to analyze and extract the unique texture features of the item surface. These features are then stored in a database for comparison during subsequent identification and verification processes. This surface feature-based identification method greatly enhances the flexibility and reliability of item identification, reduces the dependence on traditional physical markings, and provides higher anti-counterfeiting capabilities.

[0059] The present invention not only improves the accuracy and security of item identification but also has high adaptability and scalability, can be applied to a variety of different types of item identification scenarios, and provides a more reliable and efficient solution for various industries. Its application is not limited to the above fields and can also be extended to other occasions that require precise item identification and verification. For example, in medical device management, the present invention can be used to identify and track the usage of medical devices, thereby improving medical safety and management efficiency. In the field of cultural heritage protection, the present invention can be used to authenticate and protect cultural relics and prevent the forgery and loss of cultural relics.

[0060] In addition, the systems and methods of the present invention can also play an important role in fields such as food safety traceability, precious item management, and military product monitoring. In terms of food safety traceability, the unique features on the packaging can be identified to ensure the origin and quality of food and prevent counterfeit and shoddy products from entering the market. In terms of precious item management, the unique features of jewelry, artworks, etc. can be identified to ensure the authenticity and origin of these items, thereby improving management efficiency and security. In terms of military product monitoring, the unique features on weapons and equipment can be identified to ensure their traceability and security throughout the life cycle.

[0061] In summary, by combining advanced image processing technology and feature extraction technology, the present invention provides a brand-new method for item recognition and verification. This method not only improves the accuracy and security of item recognition, but also has high flexibility and adaptability, and can be widely applied to various occasions requiring item recognition, with significant practical value and broad application prospects. With the continuous development of technology and the continuous expansion of application fields, the present invention is expected to play a greater role in the future and provide more reliable and efficient item recognition solutions for all walks of life.

[0062] The intelligent item authentication system of the present invention has significant advantages in multiple application scenarios, including high efficiency, accuracy, flexibility, and cost-effectiveness. The system is applicable to the recognition of memory modules and is widely used in fields such as logistics tracking, retail management, security inspection, and production line automation, providing reliable item authentication solutions. Through technological innovation and optimization, the present invention will bring higher precision and efficiency to the field of item authentication and meet diverse application requirements.

[0063] As Figure 1-2 shown, the present invention provides an intelligent item authentication system that realizes efficient and accurate authentication and verification by extracting and analyzing the minute features on the surface of an item; it includes a data acquisition module, a preprocessing module, an authentication area selection module, a feature extraction module, a feature fingerprint processing module, a feature fingerprint matching module, an output feedback module, a database, and a user interface.

[0064] The data acquisition module is used to acquire the image data of the target item; the preprocessing module is used to perform denoising, normalization, and contrast enhancement processing on the image data; the authentication area selection module is used to automatically select the authentication area in the preprocessed image data to exclude background interference; the feature extraction module is used to identify and extract feature points within the authentication area and generate corresponding high-dimensional vectors; the feature fingerprint processing module is used to combine the high-dimensional vectors to generate a feature fingerprint; the feature fingerprint matching module is used to match the generated feature fingerprint with the feature fingerprint in the database and calculate the matching degree; the output feedback module is used to display the matching result; wherein, the matching result includes the matching degree, the number of matching pairs, and the detailed information of the matching points.

[0065] First, data of the target item is captured using, including but not limited to, a smartphone or a high-resolution imaging device, the image is acquired and preprocessed to ensure image quality. The system automatically selects the authentication area in the image to exclude background interference and ensures that only the feature part of the item is extracted. The feature extraction module uses various algorithms to identify and extract the feature points on the surface of the item, generates high-dimensional descriptors, and organizes them into a feature fingerprint for storage.

[0066] When another item needs to be verified, the system repeats the steps of taking pictures, preprocessing, selecting the authentication area, feature extraction, and generating the feature fingerprint. The newly generated feature fingerprint is matched with the fingerprints in the database, and it is judged whether the two items are the same by calculating the matching degree. The matching result is displayed through the user interface, including the matching degree, the number of matching pairs, and the detailed information of the matching points, which is convenient for further analysis and decision-making.

[0067] The working principle process of the system:

[0068] In the first step, the user first uses the data acquisition device to obtain the data of the target item. These devices must be able to provide high-resolution and high-quality data to ensure the accuracy of subsequent feature extraction. The selection of the data acquisition device should be based on the specific situation of the item and the environmental conditions to ensure the comprehensiveness and accuracy of the data.

[0069] In the second step, the data is transmitted to the feature extraction module, where the system automatically identifies and extracts the feature points in the data. These feature points are unique parts of the item surface, and through a certain algorithm, corresponding high-dimensional vectors are generated. The feature extraction module needs to have high sensitivity and intelligence to accurately extract the unique features of the item in a complex background.

[0070] In the third step, the extracted feature points and their descriptors are combined into a feature fingerprint. The feature fingerprint is a high-dimensional mathematical model that contains the information of the unique features of the item surface, similar to the "fingerprint" of the item. The generation of the feature fingerprint needs to consider the stability and uniqueness of the features to ensure accurate matching under different conditions.

[0071] In the fourth step, when it is necessary to verify whether two data sets correspond to the same item, the system calls the matching and verification module. This module compares the feature fingerprints of the two data sets and judges whether these data belong to the same item by calculating the matching degree. The calculation of the matching degree is based on the similarity between the feature points, and the system will give a clear matching result. The matching and verification module needs to have efficient and accurate computing capabilities to quickly find matching features in a large amount of data.

[0072] In the fifth step, the matching result is stored in the database and fed back to the user through the user interface. The user can view the recognition result through the interface to judge whether the items are the same. The user interface needs to provide a clear and intuitive result display and operation interface to help the user quickly understand the recognition result and perform subsequent operations.

[0073] Through the above steps, the system of the present invention can achieve fast and accurate item recognition and verification in various application scenarios. It not only improves the reliability and security of recognition, but also has high flexibility and adaptability, and can be widely applied to different fields, such as logistics tracking, retail management, security inspection, and production line automation. Whether it is rapid sorting in the logistics field, anti-counterfeiting verification in the retail field, or item recognition in security inspection, this system can provide efficient and reliable solutions. The design of the system takes into account the requirements of various application scenarios, ensuring stable operation under different environments and conditions, and providing users with efficient and convenient services.

[0074] (1) Data acquisition module

[0075] The data acquisition device is a device used to acquire item data and is responsible for obtaining high-quality data from the target item. These data will serve as the basis for subsequent feature extraction and recognition, determining the accuracy and reliability of the entire system. It can be a smartphone, an industrial camera, an X-ray machine, a near-infrared spectrometer, or other high-resolution data capture devices. This device can provide clear and detailed data under various conditions. These data are not limited to images, but also include any data that can be captured from the item, such as ultrasonic waves, infrared spectra, X-ray spectra, etc. These devices can be flexibly applied to various environments and item types to ensure the diversity and comprehensiveness of the data.

[0076] High-definition photos are one of the most commonly used data forms in the data acquisition module, mainly used for capturing the surface features of items under visible light. The device for obtaining high-definition photos can be a smartphone, an industrial camera, or other high-resolution imaging devices. Both smartphones and industrial cameras should have high-pixel sensors capable of capturing the details of the item surface. Generally speaking, the resolution of the device should be no less than 2 million pixels to ensure the clarity and richness of details of the image. The device should be able to provide high-quality images under various lighting conditions. The influence of ambient light needs to be considered to ensure that there is no overexposure or shadow during the shooting process. When necessary, lighting devices such as ring lights or LED light strips can be used to provide uniform lighting. The captured image should have no obvious blurring, noise, or other defects that affect feature extraction. The clarity of the image should be high enough to clearly show details such as the texture, color, and shape of the item surface.

[0077] To ensure high-quality data, the data acquisition module needs to meet the following requirements: First, whether it is image or spectral data, it should have high resolution and clarity to ensure that sufficient detailed information can be extracted; Second, the data should have a high signal-to-noise ratio, and the noise should be as low as possible to avoid interference with feature extraction; Third, the data acquired under different conditions should be consistent to ensure accurate feature extraction and matching in different environments; Fourth, the device should be able to remain stable during long-term use to avoid a decline in data quality caused by device aging or environmental changes.

[0078] Through the above-mentioned device and method, the data acquisition module can provide high-quality data support in various application scenarios, ensuring the accuracy and reliability of subsequent feature extraction and recognition. The flexibility and adaptability of this module enable it to be widely applied to the identification and verification of various items, providing a solid foundation for the efficient operation of the entire system.

[0079] (2) Preprocessing module

[0080] Process the acquired raw data to improve the accuracy and efficiency of feature extraction. This may include steps such as denoising, normalization, and contrast enhancement. Data preprocessing is an important link in the feature extraction process. Through effective preprocessing methods, the effect of feature extraction can be significantly improved. Denoising can use methods such as Gaussian filtering and median filtering to eliminate noise in the image. Normalization is to standardize the data to a unified range to reduce the influence caused by differences in data amplitude. Methods for enhancing contrast include contrast stretching and histogram equalization. By increasing the contrast of the image, features become more obvious and easier to detect.

[0081] (3) Authentication area selection module

[0082] Before feature extraction, select the authentication area. This can be achieved through various methods, such as Canny edge detection, Sobel operator, Laplacian operator, GrabCut, graph cut algorithm, K-means clustering, and live body cutting. The selection of the authentication area is a key step to ensure the accuracy of feature extraction. Canny edge detection, Sobel operator, and Laplacian operator are common edge detection methods. By detecting the edges and contours in the image, the boundaries of the item are determined. The GrabCut algorithm iteratively optimizes to obtain an accurate segmentation of the item by specifying an initial bounding box by the user. The graph cut algorithm is based on graph theory and divides the image into foreground and background through the minimum cut / maximum flow algorithm. K-means clustering clusters pixels into several categories based on color or other features, and selects the clustering category containing the item as the authentication area. The live body cutting method iteratively optimizes the contour through energy minimization to make it converge to the item boundary.

[0083] (4) Feature Extraction Module

[0084] The Feature Extraction Module is responsible for extracting representative feature points from the acquired item data and converting these feature points into high-dimensional vectors. This process involves a variety of algorithms and techniques, and each method has its unique advantages and applicable scenarios. This module includes a set of algorithms that can automatically identify and extract the unique features of an item and transform the features into high-dimensional vectors through a mathematical model. The flexibility and intelligence of the Feature Extraction Module enable it to process different types of data. Whether it is two-dimensional images, three-dimensional scan data, or even spectral data, it can effectively extract key features. The working principle of the Feature Extraction Module and common feature extraction methods will be introduced in detail below.

[0085] The main task of the Feature Extraction Module is to identify and extract representative feature points from the input data. These feature points can describe the uniqueness of the item and are transformed into high-dimensional vectors through a mathematical model. The whole process includes the following steps:

[0086] In the first step, feature extraction is performed on the data within the authentication area. Common feature extraction methods include Harris corner detection, Shi-Tomasi corner detection, FAST, BRIEF, FREAK, HOG, GLOH, SIFT, SURF, etc. The selection of feature extraction methods should be based on the specific application scenario and data type. Harris corner detection detects significant features such as corners by calculating the image gradient matrix. Shi-Tomasi corner detection improves the Harris algorithm and determines the significance of corners through eigenvalue judgment. FAST detects corners by quickly testing the pixel neighborhood and has high computational efficiency. BRIEF generates binary descriptors by performing binary comparisons on pixel pairs in the neighborhood of feature points. FREAK designs feature descriptors based on the structure of the human retina and uses a circular structure for feature extraction. HOG describes local features by calculating the direction histogram of image gradients and is widely used in object detection and recognition. GLOH is an improved version of HOG and describes features through a more detailed gradient direction histogram. SIFT is a scale-invariant feature extraction method that can detect and describe local features in images, has good anti-rotation and scale change performance, and is suitable for image matching and object recognition in various scenarios. SURF is an accelerated version of SIFT that uses the concept of integral images to accelerate the process of feature point detection and description, can improve computational efficiency while maintaining performance, and is suitable for real-time processing.

[0087] In the feature point detection stage, various algorithms determine significant feature points in the data according to different criteria. For example, the Harris corner detection algorithm detects corner points by calculating the image gradient matrix. The detection of feature points is a crucial step in the feature extraction process, and different algorithms have their own advantages in terms of the stability and robustness of detecting feature points. Feature point detection algorithms based on traditional methods, such as Harris and Shi-Tomasi, are simple and fast to calculate and are suitable for application scenarios with limited resources.

[0088] In the second step, high-dimensional vectors are generated. The process of generating high-dimensional vectors is to convert the local information of each feature point into a high-dimensional vector, which is called a descriptor. The generation of descriptors involves a series of mathematical operations, such as gradient calculation, direction assignment, and histogram calculation. The generation of high-dimensional vectors is one of the key steps in the feature extraction module. The quality of the descriptor directly affects the subsequent matching and recognition effects, so the generation process of the descriptor needs to be carefully designed and optimized. Gradient calculation is the basis for generating descriptors. By calculating the gradient direction and amplitude of the pixels around each feature point, local direction information is obtained. Direction assignment assigns a direction and weight to each pixel according to the gradient direction and amplitude. Histogram calculation divides the local area into several sub-areas, statistically analyzes the gradient directions in each sub-area, and generates a direction histogram. Finally, the direction histograms of all sub-areas are combined into a high-dimensional vector, which is the descriptor of the feature point.

[0089] Through the above steps, the feature extraction module can extract stable and representative high-dimensional vectors from the input data, and these vectors will serve as the basis for subsequent matching and recognition. The design and implementation of the feature extraction module determine the performance and reliability of the entire system. Therefore, it is necessary to comprehensively consider the advantages and disadvantages of various algorithms and select the most suitable feature extraction method according to the specific application scenario. The core of the feature extraction module lies in how to efficiently and accurately extract representative feature points from a large amount of data and generate high-dimensional vectors that can stably represent the features of items. This not only requires the algorithm itself to be efficient and robust but also needs to be continuously optimized and adjusted in practical applications to adapt to different items and data environments.

[0090] (5) Feature fingerprint processing module

[0091] The feature fingerprint processing module combines the high-dimensional vectors corresponding to the extracted feature points into an overall feature fingerprint. This fingerprint can be regarded as the unique identifier of the item and is used for subsequent recognition and verification. The feature fingerprint generation module not only needs to consider the extraction of features but also ensure that these features can stably and uniquely represent the item, so as to accurately identify it under different conditions.

[0092] The formation of the feature fingerprint is a crucial step after the feature extraction module. This process organizes the extracted feature point information into a format that the system can process, for subsequent matching and verification. The following will detail the steps of forming the feature fingerprint and the storage method of the feature fingerprint. The formation of the feature fingerprint includes several important steps, ensuring the integrity and availability of the feature information.

[0093] In the first step, the entire image is divided into several small blocks. The purpose of this step is to subdivide the image, making the feature extraction more precise and localized. The way of block division can be adjusted according to the size of the image and the required precision. By reasonable block division, the accuracy and efficiency of feature extraction can be improved. This subdivision strategy can ensure that each image block can be fully analyzed, avoiding information loss or errors that may be caused by overall analysis, thus improving the robustness and reliability of the entire system.

[0094] In the second step, within each image block, several feature points are extracted using the aforementioned feature extraction methods (such as Harris corner detection, FAST, etc.). These feature points are representative local features within the image block, capable of reflecting the uniqueness of the image block. The selection and extraction of feature points are the basis for ensuring the accuracy of subsequent processing, and different feature extraction methods can be selected according to specific application scenarios. For example, for some application scenarios with high real-time requirements, the FAST algorithm with higher computational efficiency may be selected.

[0095] In the third step, for each feature point, a corresponding descriptor is generated. The descriptor is a high-dimensional vector representation of the feature point, containing various information of the feature point, such as gradient, direction, etc. The process of generating the descriptor is through a series of mathematical operations to convert the information of the feature point into a high-dimensional vector, and these vectors can comprehensively describe the attributes of the feature point. The generation of the descriptor is the core step in the feature extraction process. Through the descriptor, the complex information of the feature point can be simplified into a high-dimensional vector that can be quantified and compared, making the subsequent matching and verification more efficient and accurate.

[0096] In the fourth step, the index of each image block and the descriptors of all feature points within that block are combined into a matrix. This step combines the position information and feature information of the image block to form a complete representation. This combination not only retains the detailed information of the feature points but also records their positions in the image, making subsequent processing more efficient and accurate. By combining the image block index and the descriptor matrix, the integrity and consistency of the feature information are ensured, providing a reliable data basis for subsequent matching algorithms.

[0097] Step 5: Write the index and descriptor matrix of each image patch as a key-value pair. The structure of the key-value pair can be designed according to specific requirements. Generally, the key is the index of the image patch, and the value is the descriptor matrix of the feature points within the patch. Combine the key-value pairs of all image patches into a dictionary. This dictionary contains the feature information of the entire image and is a complete representation of the image. In this way, the local features of the image can be systematically and structurally organized, making the subsequent matching and verification processes more convenient and efficient. The application of the dictionary structure not only improves the efficiency of data organization but also facilitates subsequent data retrieval and processing, ensuring that the system can maintain high performance in large-scale data processing.

[0098] Step 6: Store the generated feature fingerprint dictionary in an appropriate format. Commonly used storage formats include.npz, etc. These formats can efficiently store and read large-scale data. This storage format can provide fast access and retrieval capabilities while ensuring data integrity. To ensure data security and reliability, the feature fingerprint data can also be compressed and optimized to reduce storage space and improve the reading speed. In addition, the feature fingerprint data can be encrypted and backed up as needed to prevent data loss or unauthorized access. In terms of storage paths and management, an indexing system can be established to classify and store the feature fingerprint data for subsequent retrieval and use. The establishment of the indexing system can greatly improve the efficiency of data retrieval and ensure that the corresponding feature fingerprint data can be quickly found when matching and verification are required.

[0099] Through the above detailed steps and methods, the formation and storage process of the feature fingerprint are completed. As the unique identifier of an item, the feature fingerprint can play an important role in subsequent matching and verification. The design and implementation of the feature fingerprint need to consider not only the accuracy of feature extraction but also the efficiency of storage and retrieval to ensure the performance and reliability of the entire system. In practical applications, the formation and storage of the feature fingerprint need to be optimized and adjusted according to specific application scenarios to meet different requirements and conditions. Specifically, in actual deployment, the size of image patches and the feature extraction method can be flexibly adjusted according to different application requirements and scenario characteristics to achieve the best recognition effect. At the same time, according to the actual storage and computing resources, the feature fingerprint data can be appropriately compressed and optimized to ensure the efficiency and stability of the system during actual operation. In short, through scientific and reasonable design and optimization, the formation and storage of the feature fingerprint can not only meet the current recognition needs but also have strong scalability and adaptability to adapt to future changing and developing application requirements.

[0100] (6) Feature fingerprint matching module

[0101] The feature fingerprint matching module is used to compare the feature fingerprints in different data, and determine whether these data belong to the same item by calculating the matching degree. This module includes a matching algorithm that can effectively calculate the matching result. The matching and verification module needs to have efficient and accurate algorithm capabilities, be able to quickly find matching features in massive data, and ensure the timeliness and accuracy of identification. The following will introduce in detail the method of matching feature points and the process of calculating the matching result:

[0102] After the feature points are extracted and the feature fingerprints are formed, the matching algorithm needs to compare the feature points in different images to determine whether they belong to the same item. Commonly used methods for matching feature points include brute-force matching, K-nearest neighbor matching, bidirectional matching, RANSAC, etc. The brute-force matching method finds the matching pair with the smallest distance by calculating the distance between each feature point descriptor and all other feature point descriptors. Its principle is simple, but the computational cost is large, and it is suitable for cases with fewer feature points. The KNN matching method finds K feature points with the closest distance for each feature point, and then selects the best match based on the distance. By calculating the K nearest neighbor distances of each feature point and selecting the one with the smallest distance as the matching point, the optimal match can be further screened by combining the distance ratio (such as 0.8). The matching accuracy is high, but the computational complexity is also high. The bidirectional matching method requires that the matching must be bidirectional when performing matching, that is, the feature points of image A match the feature points of image B, and at the same time, the feature points of image B also match the feature points of image A. Each feature point is verified for bidirectional matching, and only the matching pairs that meet the conditions in both directions are considered valid matches. This method reduces false matches and improves the accuracy of matching, but the computational amount increases. The RANSAC algorithm identifies and removes false matches by randomly sampling feature points and calculating model parameters. Randomly select a part of the matching points, calculate the model parameters (such as the homography matrix), verify the compliance of the remaining points, and iterate multiple times to find the best model. RANSAC has strong anti-interference ability and can effectively remove false matches, but the computational complexity is high, and the number of iterations and the inlier threshold need to be set.

[0103] After the feature point matching is completed, it is necessary to calculate the matching result to determine whether two images belong to the same item. The process of calculating the matching result includes the following steps. First, according to the matching method, effective matching pairs are screened out. Ways such as distance threshold, two-way verification, and block index alignment can be combined to ensure the accuracy and reliability of the matching. The screening of matching pairs is a key step to ensure the matching quality. By strict screening criteria, the accuracy of the matching can be significantly improved. To screen the matching pairs, it is necessary to calculate the distance between each pair of matching feature points. Commonly used distance metrics include Euclidean distance and Hamming distance. Euclidean distance is applicable to the case where the descriptor is a floating point number, while Hamming distance is often used for binary descriptors. Distance calculation is the basis of the matching result. By calculating the distance between each pair of feature points, their similarity can be quantified. Matching point pairs with a distance less than a pre-set threshold are called effective matching pairs. Then, count the number of effective matching pairs to obtain an overall matching metric value. The number of effective matching pairs reflects the overall similarity of the matching. Simple threshold judgment can be used, or more statistical methods such as standard deviation and the number of matching pairs can be combined. The calculation of the matching degree not only depends on the average distance, but other statistics can also be considered to improve the accuracy and robustness of the judgment.

[0104] (7) Output feedback module

[0105] The output feedback module is used to output the matching result, including the number of effective matching pairs, detailed information of the matching points, etc. The output of the matching result needs to be detailed and comprehensive to facilitate further analysis and verification by the user. The number of effective matching pairs reflects the overall similarity of the images, and the detailed information of the matching points provides more specific matching situations. Through the above steps, the matching algorithm can effectively compare the feature fingerprints in different images, calculate the matching result, and output the final matching conclusion. It not only ensures the accuracy and reliability of the matching, but also provides flexibility and adaptability to meet the requirements of different application scenarios. In practical applications, the selection and optimization of the matching algorithm need to be adjusted according to the specific application scenario and data characteristics to achieve the best matching effect. Different matching algorithms and parameter settings can be optimized according to the actual situation to ensure that the system can maintain efficient and accurate performance under various conditions.

[0106] The reliability and efficiency of the matching algorithm directly affect the performance of the entire system. Therefore, when selecting and optimizing the matching algorithm, various factors need to be fully considered, including the type and scale of data, the limitations of computing resources, and the specific application requirements. Through reasonable algorithm design and optimization, while ensuring the accuracy of matching, the processing speed and efficiency of the system can be improved. In addition, in practical applications, the matching algorithm also needs to have a certain degree of robustness to cope with various noises and interferences, ensuring high-efficiency and stable performance in complex environments. To further improve the performance of the matching algorithm, multiple matching methods can be combined, making comprehensive use of their advantages to construct a hybrid matching model to achieve higher matching accuracy and efficiency. In short, through scientific and reasonable algorithm design and optimization, the matching algorithm can play an important role in various application scenarios, ensuring the high efficiency and reliability of the system.

[0107] (8) Database

[0108] The database is used to store the feature fingerprint information of items. Each identified item has its feature fingerprint stored in the database for future comparison and verification. The database not only needs to store a large amount of feature fingerprint data but also provide functions for fast retrieval and comparison to ensure the system remains efficient when processing a large number of requests.

[0109] (9) User Interface

[0110] The user interface provides an intuitive interface for users to perform operations and view recognition results. The user interface can be an application or a web platform, facilitating users to upload data, perform recognition operations, and view results. The design of the user interface should focus on the user experience, providing a simple and intuitive operation process to help users quickly understand and use the system functions.

[0111] As Figure 3 shown, this embodiment demonstrates how to use a smartphone to take a photo of a memory module, preprocess the image, select the authentication area, extract features, generate and save the feature fingerprint through the intelligent vision similar item recognition system, then perform the same operations on another memory module, and finally perform matching and verification. First, use a smartphone to take a photo of a memory module. The obtained image may contain noise and other unnecessary information, so preprocessing is required. The preprocessing steps include denoising, normalization, and enhancing contrast to ensure that the image quality meets the requirements of feature extraction. These steps can effectively improve the clarity and contrast of the image, making the subsequent feature extraction process more accurate and reliable.

[0112] In the preprocessed image, the system automatically selects the authentication area. The selection of the authentication area can be achieved through various methods, such as algorithms like contour detection. The purpose of selecting the authentication area is to ensure that only the characteristic parts of the memory module are extracted, excluding background interference. This process ensures that the system only focuses on the memory module itself and is not affected by the surrounding environment, thereby improving the accuracy of recognition. When selecting the authentication area, the system will perform multiple iterations of optimization to find the best characteristic area. This not only improves the efficiency of feature extraction but also enhances the system's adaptability to images under different lighting and angle conditions.

[0113] Next, feature extraction is performed on the image within the authentication area. Feature extraction algorithms (such as Harris corner detection, FAST, etc.) are used to identify and extract the feature points on the surface of the memory module. Each feature point generates a high-dimensional descriptor, which contains various information about the feature point, such as gradient, direction, etc. The process of feature extraction is the core step of the intelligent vision similar item recognition system. By extracting the detailed information of each feature point, the system can comprehensively describe the characteristics of the memory module. The selection and parameter optimization of the feature extraction algorithm are crucial. Through continuous debugging and optimization, the system can maintain a high-precision feature extraction ability in different environments.

[0114] Then, the extracted feature points and descriptors are organized into a feature fingerprint. Specifically, the authentication area is divided into several small blocks. Feature points are extracted within each small block, and a descriptor matrix is generated. The index of each small block and the descriptor matrix are formed into key-value pairs, and all key-value pairs form a dictionary to form a feature fingerprint. Finally, the feature fingerprint is stored in the.npz format. The generation process of the feature fingerprint ensures that the characteristic information of each memory module can be completely recorded and stored, facilitating subsequent matching and verification. Through this structured storage method, the system can quickly retrieve and compare feature fingerprints, improving the matching efficiency.

[0115] Use a smartphone to take a photo of another memory module and repeat the above steps (preprocessing, authentication area selection, feature extraction, feature fingerprint generation). The newly generated feature fingerprint is matched with the existing feature fingerprints in the database. Matching algorithms (such as brute-force matching, K-nearest neighbor matching, bidirectional matching, RANSAC, etc.) are used to compare the two feature fingerprints, calculate the distance between each pair of feature points, and filter out valid matching pairs. This step is an important link to ensure the accuracy of matching. Through a reasonable matching algorithm, the accuracy and reliability of matching can be improved. The selection of the matching algorithm depends on the specific application scenario. For scenarios with high real-time requirements, an algorithm with higher computational efficiency can be selected, while for scenarios with high accuracy requirements, a more complex matching algorithm can be selected.

[0116] Based on the matched feature point pairs, the system calculates the matching result. First, it calculates the distance between each pair of matched feature points. Commonly used distance metrics include Euclidean distance and Hamming distance. Then, it calculates the number of valid matching pairs that meet the conditions. Based on the number of valid matching pairs and a preset threshold, the matching result is calculated. These calculation steps are an important part of ensuring the accuracy of the matching result. Through precise calculation methods, the similarity between images can be effectively quantified. The system can also perform further statistical analysis based on the distribution of the matching degrees to improve the reliability of the matching result.

[0117] The system determines whether two images belong to the same memory module. If the number of valid matching pairs is higher than the preset threshold, the system determines that they belong to the same memory module; otherwise, the system determines that they belong to different memory modules. This determination process ensures that the system can accurately distinguish different memory modules and avoid misjudgment. The threshold of the system can be adjusted according to the actual application situation to achieve the best recognition effect. By dynamically adjusting the threshold, the system can maintain high efficiency and stable performance in different environments and application scenarios.

[0118] The system outputs the matching result to the user. The matching result includes the number of valid matching pairs, detailed information of the matching points, etc. The user can view the matching result through the system interface and perform operations and make decisions according to needs. The system interface is designed to be simple and clear, and the operation is convenient. The user can easily get started and perform operations. Through the detailed and comprehensive display of the matching result, the user can clearly understand the specific situation of each recognition and matching. The system interface also provides a variety of data visualization tools to help the user intuitively understand and analyze the matching result, improving the user experience and operation efficiency.

[0119] Through the above steps, the intelligent vision similar item recognition system can efficiently and accurately identify and verify memory modules. The system not only improves the accuracy and efficiency of identification but also ensures the simplicity of operation and the user experience. This embodiment demonstrates the operation process and effect of the intelligent vision similar item recognition system in actual applications, providing a reference for other application scenarios. Through this scientific and reasonable implementation method, the intelligent vision similar item recognition system can play an important role in various application scenarios, ensuring the high efficiency and reliability of the system and meeting the needs of different users. In the future, the intelligent vision similar item recognition system can further expand its functions and optimize its performance to provide efficient and reliable item identification solutions for more fields.

[0120] The intelligent vision similar item recognition system is not only applicable to the recognition of memory modules, but also can be widely used in other fields, such as cultural relics identification, product anti-counterfeiting, asset management, etc. The flexibility and adaptability of the system enable it to meet the diverse needs of different users. Through continuous technological innovation and optimization, the intelligent vision similar item recognition system will play its unique advantages in more fields and provide users with more accurate and efficient services. In short, relying on its excellent performance and broad application prospects, the intelligent vision similar item recognition system will surely play an increasingly important role in future development and become one of the key technologies in the field of item identification.

[0121] The advantages of the present invention include:

[0122] The intelligent vision similar item recognition system has wide applicability and remarkable advantages in many application scenarios. Its main application scenarios include logistics tracking, retail management, security inspection, production line automation, etc. In these application scenarios, the system improves work efficiency and security, and reduces labor costs and error rates through efficient and accurate item identification and verification.

[0123] 1> In logistics tracking, the intelligent vision similar item recognition system can accurately track items in each logistics link, ensuring the safety and integrity of goods during transportation. By monitoring and identifying goods in real time, it can effectively prevent goods from being lost or stolen, and improve the transparency and reliability of logistics management. Especially in international logistics, the intelligent vision similar item recognition system can ensure the safety and timeliness of goods in cross-border transportation through global positioning and real-time data update, reducing logistics delays and losses.

[0124] 2> In retail management, the intelligent vision similar item recognition system can be used in inventory management, product anti-counterfeiting and customer service, etc. By quickly identifying and verifying products, the system can help merchants conduct accurate inventory management, reducing inventory backlogs and out-of-stock situations. At the same time, the system can also identify counterfeit and shoddy products, protect brand reputation, and improve the shopping experience of consumers. For high-value products, such as luxury goods and jewelry, the intelligent vision similar item recognition system can provide detailed appraisal reports, enhancing consumers' purchase confidence and the market competitiveness of the brand.

[0125] 3> In security inspection, the intelligent vision similar item recognition system is widely used in security inspection work at airports, stations, customs and other places. The system can efficiently identify dangerous items in luggage and goods, improving the efficiency and accuracy of security inspection and ensuring public safety. In addition, the system can also be used to verify the identities of personnel and vehicles in border inspections, effectively preventing illegal entry and smuggling activities. Especially in large-scale public events and conferences, the intelligent vision similar item recognition system can quickly screen the items carried by participants to ensure the safety and smooth progress of the event.

[0126] 4> In production line automation, intelligent vision similar item recognition systems can be used for quality control, production monitoring, equipment maintenance, etc. By identifying and verifying products on the production line in real time, the system can promptly detect and correct problems in the production process, ensuring product quality. At the same time, the system can also monitor the operating status of production equipment, perform maintenance and upkeep in a timely manner, and avoid equipment failures and production stoppages. Intelligent vision similar item recognition systems can also be integrated into production management systems to achieve automatic collection and analysis of production data, improving the intelligence and automation level of the production line.

[0127] Intelligent vision similar item recognition systems have significant advantages compared to traditional anti-counterfeiting methods.

[0128] 1> First, compared with traditional anti-counterfeiting methods such as QR codes, barcodes, and anti-counterfeiting labels, intelligent vision similar item recognition systems are not easily tampered with or transferred. These traditional methods usually rely on physical labels, which are easy to copy, damage, or remove, presenting certain security risks. In contrast, intelligent vision similar item recognition systems identify by extracting minute features on the surface of items, do not require additional physical labels, do not affect the appearance of the items, and are difficult to copy and tamper with, offering higher security.

[0129] 2> Second, compared with technologies such as RFID and NFC, intelligent vision similar item recognition systems are less costly. RFID and NFC technologies require specific tags and reading / writing devices, with high costs and being difficult to deploy in some application scenarios. In contrast, intelligent vision similar item recognition systems only require common image acquisition devices such as smartphones and industrial cameras, with lower costs, more flexible deployment, and a wider scope of application. The high cost performance of the system makes it more competitive in large-scale applications.

[0130] The advantages of intelligent vision similar item recognition systems are mainly reflected in the following aspects. First, the system is efficient and accurate. Through advanced feature extraction and matching algorithms, the system can quickly and accurately identify and verify items, reducing the time and error rate of manual operations. Second, the system has a high degree of flexibility and adaptability. The system can be customized and optimized according to different application scenarios and requirements, and is suitable for various complex working environments. For example, in a complex production environment, the system can be customized according to the specific situation of the production line to ensure the best performance of the system. In addition, the system also has good scalability. With the development of technology and changes in application requirements, the system can be continuously upgraded and expanded to meet future application needs. The modular design of the system enables it to easily integrate new functions and technologies to adapt to the ever-changing market demands.

[0131] The intelligent vision similar item recognition system also has high security and reliability. The system ensures the accuracy and reliability of the identification results through strict matching and verification processes. At the same time, the system adopts advanced data encryption and storage technologies to protect the security of user data. For sensitive data, the system can implement multi-level security protection measures to prevent data leakage and illegal access. In addition, the system also has a good user experience. The system interface is designed to be simple and clear, and the operation is convenient, so that users can easily get started and operate.

[0132] In summary, the intelligent vision similar item recognition system demonstrates powerful functions and remarkable advantages in multiple application scenarios. Through efficient and accurate identification and verification, the system not only improves work efficiency and security, but also reduces costs and enhances the user experience. In the future, with the continuous progress of technology, the intelligent vision similar item recognition system will play an important role in more fields, bringing more application value and social benefits. For example, in the construction of smart cities, the intelligent vision similar item recognition system can be used in aspects such as urban management, public safety, and environmental monitoring to improve the intelligent level of urban management and improve the quality of life of residents. Through continuous innovation and optimization, the intelligent vision similar item recognition system will make greater contributions to social development and economic growth.

[0133] As Figure 4-5 shown, the present invention also provides an intelligent item identification method. Based on the above-mentioned intelligent item identification system, the data of the item is converted into a feature fingerprint that can be used for identification, and the identification and verification of the item are realized through a matching algorithm. Specifically, it includes the following steps:

[0134] S1. Use an image acquisition device to acquire the original image data to be recognized and verified. The original image data can be acquired through various data acquisition devices (including but not limited to smartphones, industrial cameras, ultrasonic detectors, infrared spectrometers, etc.). The acquired images may need to be preprocessed to improve the accuracy of subsequent feature extraction and matching.

[0135] S2. Preprocess the initial image data to obtain the initial image data, which can help improve the image quality and make feature extraction more accurate. Among them, the preprocessing includes denoising, normalization, and enhancing contrast.

[0136] S3. Automatically select the authentication area in the initial image data to exclude background interference;

[0137] S4. In the authentication area, use Harris corner detection and FAST to identify and extract feature points, and generate corresponding high-dimensional descriptors, as Figure 6 shown.

[0138] After preprocessing the image and selecting the authentication area, feature extraction is performed on the preprocessed image. The feature extraction module will identify and extract representative feature points from the image and generate corresponding high-dimensional descriptors. The methods of feature extraction can be Harris corner detection, FAST, etc. The result of this step is the feature fingerprint of each image, which contains the detailed feature information of the image, and these feature information will be used as the basis for subsequent matching.

[0139] S5. Divide the initial image into several image blocks, extract feature points within each image block, generate a descriptor matrix, and form key-value pairs by combining the index of each image block and the descriptor matrix, so as to form a dictionary with all key-value pairs to form a feature fingerprint, as Figure 7 shown.

[0140] Specifically, divide the image into several small blocks, extract feature points within each small block, and generate a descriptor matrix. Then, form key-value pairs by combining the index of each small block and the descriptor matrix, and all key-value pairs form a dictionary to form a feature fingerprint. This method not only retains the detailed information of the feature points but also records their positions in the image, making subsequent processing more efficient and accurate.

[0141] S6. Compare the feature points of the feature fingerprint with the feature fingerprints in the database using a matching algorithm, screen out valid matching pairs, and calculate the number of valid matching pairs to obtain an overall matching metric value as the matching degree.

[0142] After the feature fingerprint is formed, the feature fingerprints of the image to be recognized and verified will be compared. Use a matching algorithm (such as brute-force matching, K-nearest neighbor matching, bidirectional matching, RANSAC, etc.) to compare the feature points of the two images and find the matching feature point pairs. The matching algorithm will calculate the distance between each pair of feature points and screen out valid matching pairs. This step is an important link to ensure the accuracy of the matching. By using a reasonable matching algorithm, the accuracy and reliability of the matching can be improved.

[0143] According to the matching feature point pairs, calculate the matching result. First, calculate the distance between each pair of matching feature points. Commonly used distance metrics include Euclidean distance and Hamming distance. Then, compare the distance of each pair of matching pairs with a threshold to screen out valid matching pairs, and obtain an overall matching metric value by calculating the number of valid matching pairs. Finally, calculate the matching degree according to the average matching distance and the preset threshold. These calculation steps are an important part to ensure the accuracy of the matching result. By using precise calculation methods, the similarity between images can be effectively quantified.

[0144] S7. If the matching degree is lower than the preset threshold, it is determined that the two images belong to the same item; otherwise, it is determined that they belong to different items, and finally the matching result is output, including the matching degree, the number of matching pairs, and the detailed information of the matching points.

[0145] To determine whether two images belong to the same item, it is based on the matching result. According to the calculated matching degree, it is compared with the preset threshold. If the matching degree is lower than the threshold, it is determined that the two images belong to the same item; otherwise, it is determined that they belong to different items. The matching result includes the number of valid matching pairs, the detailed information of the matching points, etc. These results will provide users with detailed and comprehensive information to help make further judgments and decisions.

[0146] Outputting the matching result to the user is the last step. The matching result needs to be detailed and comprehensive, including the number of matching pairs, the detailed information of the matching points, etc., so that users can conduct further analysis and verification. Users can view the matching result through the system interface and perform operations and make decisions according to their needs. The design of the system interface needs to be simple and clear to ensure that users can easily view and understand the matching result.

[0147] The judgment criterion of the matching result is the key to ensuring that the system can accurately identify and verify items. The number of valid matching points is an important indicator to measure the similarity between two images. The system will preset a threshold. When the number of valid matching pairs is lower than this threshold, it is determined that the two images do not belong to the same item; otherwise, it is determined that they belong to the same item.

[0148] In addition to the number of matching pairs, the detailed information of the matching points also needs to be considered. It includes the specific positions of each pair of matching points, the matching distance, etc. By analyzing the distribution of the matching points, the accuracy of the matching result can be further verified. For example, if all the matching points are concentrated in a certain local area, its reliability may need to be further verified. To improve the accuracy of the matching result, more statistical methods can be combined for analysis. For example, calculate the standard deviation of the matching distance, the distribution of the number of matching pairs, etc. By comprehensively analyzing various statistics, it can be more accurately judged whether two images belong to the same item.

[0149] Through the above processes and judgment criteria, the item recognition and verification system can efficiently and accurately determine whether two images belong to the same item. This process not only ensures the accuracy of recognition but also provides flexibility and adaptability to meet the requirements of different application scenarios. In practical applications, the processes and judgment criteria for item recognition and verification can be optimized and adjusted according to specific application requirements to achieve the best recognition effect. For example, in some application scenarios with high-precision requirements, the recognition accuracy and reliability of the system can be improved by adjusting the parameters of the matching algorithm, increasing data preprocessing steps, etc. In short, through scientific and reasonable process design and judgment criteria, the item recognition and verification system can play an important role in various application scenarios, ensuring the efficiency and reliability of the system.

[0150] The present invention can accurately and reliably identify items, and can precisely capture and recognize the subtle feature differences on the surface and / or inside of items, thus effectively addressing the problem of the widespread prevalence of counterfeit and shoddy products in the current market. Through high-resolution image capture technology and advanced feature extraction algorithms, the system can analyze and extract the unique features of items, which may include but are not limited to surface texture, color, shape, and other subtle physical differences. This feature information is then stored in a secure and easily accessible database. When it is necessary to verify the authenticity of a commodity, simply capture an image again through a camera and compare it with the records in the database to quickly and accurately confirm the identity of the commodity.

[0151] The intelligent identification system of the present invention not only improves the accuracy and security of recognition but also has high adaptability and scalability, and can be applied to the item recognition needs of multiple industries and different types. For example, in medical device management, the system can be used to ensure the legality and safety of medical devices and prevent the use of counterfeit products. In the field of cultural heritage protection, the system can help identify and protect cultural relics to ensure their authenticity and the legality of their origin. In food safety tracing, by identifying the unique features on the packaging, the source and quality of food can be ensured, and counterfeit and shoddy products can be prevented from entering the market. In addition, the system can also be used for the management of valuable items such as jewelry and artworks. By identifying their unique surface features, their authenticity and origin can be ensured, thereby improving management efficiency and security. In the monitoring of military products, the traceability and security of weapons and equipment throughout their life cycle can also be ensured by identifying their unique features.

[0152] In summary, in the context of globalization, the problem of counterfeit and shoddy products is becoming increasingly severe. This invention serves as an advanced and reliable identification method to protect the interests of enterprises and consumers. With the continuous development of technology and the expansion of application fields, this invention is expected to play a greater role in the future, providing more reliable and efficient item identification and authentication services for all walks of life, thereby effectively curbing the spread of counterfeit and shoddy products, maintaining market order and social integrity. By improving the accuracy and reliability of identification technology, economic losses caused by counterfeit and shoddy products can be significantly reduced, consumer trust can be enhanced, and the market competitiveness of brands can be strengthened. This is not only beneficial to enterprises but also contributes to the healthy development of the overall economic environment.

[0153] As Figure 8 shown, the present invention also provides a computer device, which can be a server, and its internal structure can be as Figure 8 shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store all the data required for the process of the intelligent item identification method. The network interface of the computer device is used to communicate with external terminals through a network connection. The computer program, when executed by the processor, implements the intelligent item identification method.

[0154] Those skilled in the art can understand that Figure 8 the structure shown in

[0155] is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied.

[0155] An embodiment of this application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements any one of the above intelligent item identification methods.

[0156] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided in this application and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0157] It should be noted that in this article, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that includes a series of elements includes not only those elements but also other elements not expressly listed, or elements that are inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the phrase "including one..." does not exclude the presence of additional identical elements in the process, apparatus, article, or method that includes the element.

[0158] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structural or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. An intelligent item identification system, characterized in that, It includes a data acquisition module, a preprocessing module, an authentication area selection module, a feature extraction module, a feature fingerprint processing module, a feature fingerprint matching module, an output feedback module, and a database; The data acquisition module is used to acquire image data of a target item; The preprocessing module is used to perform denoising, normalization, and contrast enhancement processing on the image data; The authentication area selection module is used to automatically select the authentication area in the preprocessed image data to exclude background interference; The feature extraction module is used to identify and extract feature points within the authentication area and generate corresponding high-dimensional vectors; The feature fingerprint processing module is used to combine the high-dimensional vectors to generate a feature fingerprint; The feature fingerprint matching module is used to match the generated feature fingerprint with the feature fingerprint in the database and calculate the matching degree; The output feedback module is used to display the matching result; wherein, the matching result includes the matching degree, the number of matching pairs, and detailed matching point information.

2. The intelligent article authentication system according to claim 1, wherein The data acquisition module includes a smart phone, an industrial camera, an X-ray machine, a near-infrared spectrometer, or other data capture devices.

3. The intelligent item authentication system according to claim 1, wherein, In the preprocessing module, Gaussian filtering and median filtering methods are used for denoising processing to eliminate noise in the image, and contrast stretching and histogram equalization methods are used for contrast enhancement processing to improve the contrast of the image data.

4. The intelligent item identification system according to claim 1, characterized in that, In the authentication area selection module, Canny edge detection, Sobel operator, and Laplacian operator are used to detect the edges and contours of the preprocessed image data to determine the boundary of the target item; The GrabCut algorithm is used to iteratively optimize to obtain the precise segmentation of the target item by specifying an initial bounding box by the user; The graph cut algorithm is used to segment the image data into foreground and background based on graph theory through the minimum cut / maximum flow algorithm; The K-means clustering algorithm is used to cluster pixels into several categories by color, and the clustering category containing the target item is selected as the authentication area; The live cutting method is used to iteratively optimize the contour to converge to the boundary of the target item by minimizing energy.

5. The intelligent article authentication system according to claim 1, wherein In the feature extraction module, Harris corner detection, Shi-Tomasi corner detection, FAST, BRIEF, FREAK, HOG, GLOH, SIFT, and SURF are used to identify and extract feature points within the authentication area; Generating the corresponding high-dimensional vectors includes: Gradient calculation, obtaining local direction information by calculating the gradient direction and amplitude of pixels around each feature point; Direction assignment, assigning a direction and weight to each pixel according to the gradient direction and amplitude; Histogram calculation, dividing the local area into several sub-areas, statistically analyzing the gradient directions within each sub-area to generate a direction histogram, and combining the direction histograms of all sub-areas into a high-dimensional vector, and the high-dimensional vector is the descriptor of the feature point.

6. The intelligent item authentication system according to claim 5, characterized in that, Specifically included in the feature fingerprint processing module: Dividing the image data into several image blocks, and using the feature extraction method in the feature extraction module to extract several feature points and generate corresponding high-dimensional vectors; Combine the index of each image patch and the descriptors of all feature points within the image patch into a matrix, and write the index of each image patch and the descriptor matrix as a key-value pair; where the key is the index of the image patch and the value is the descriptor matrix of the feature points within the image patch. Combine the key-value pairs of all image patches into a dictionary as the feature fingerprint dictionary, store the feature fingerprint dictionary, and establish an indexing system to classify and store the feature fingerprint data; where the storage format is.npz.

7. The intelligent item identification system according to claim 1, wherein Specifically included in the feature fingerprint matching module are: Compare the generated feature fingerprints with the feature points of the feature fingerprints in the database using the method of matching feature points; where the method of matching feature points includes brute-force matching, K-nearest neighbor matching, two-way matching, and RANSAC. Filter out valid matching pairs according to the method of matching feature points in combination with the distance threshold, two-way verification, and block index alignment, and calculate the number of valid matching pairs to obtain the overall matching metric value.

8. An intelligent item identification method, characterized in that, Based on the intelligent item identification system according to any one of claims 1-7, comprising the following steps: Use an image acquisition device to acquire the original image data to be recognized and verified. Preprocess the initial image data to obtain the initial image data; where the preprocessing includes denoising, normalization, and enhancing the contrast. Automatically select the authentication area in the initial image data to exclude background interference. Use Harris corner detection and FAST recognition in the authentication area to extract feature points and generate corresponding high-dimensional descriptors. Divide the initial image into several image patches, extract feature points within each image patch, generate a descriptor matrix, and form a key-value pair with the index of each image patch and the descriptor matrix, so as to form a dictionary with all key-value pairs to form a feature fingerprint. Compare the feature points of the feature fingerprint with the feature fingerprints in the database using a matching algorithm, filter out valid matching pairs, and calculate the number of valid matching pairs to obtain the overall matching metric value as the matching degree. If the matching degree is lower than a preset threshold, it is determined that the two images belong to the same item; otherwise, it is determined that they belong to different items, and finally the matching result is output, including the matching degree, the number of matching pairs, and the detailed information of the matching points.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method described in claim 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method described in claim 8 are implemented.

Images