Multi-region double-stage RMB true and false identification method and electronic equipment
By employing a multi-region, two-stage identification method, combined with feature analysis and deep learning models, the problem of high resource consumption in existing banknote anti-counterfeiting technologies has been solved, achieving efficient and accurate banknote authentication.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-14
- Publication Date
- 2026-03-13
AI Technical Summary
Existing banknote anti-counterfeiting detection technologies require large datasets and computing resources, and model training and optimization are time-consuming and labor-intensive, making it difficult to meet actual production needs.
A multi-region, two-stage identification method is adopted. By acquiring ticket sample image information for correction and registration, feature values of the anti-counterfeiting identification area are extracted and judged by combining a classification network model, and secondary confirmation is performed by combining feature analysis and deep learning model.
It improves the accuracy and efficiency of banknote authentication, reduces the demand for computing resources, and meets actual production needs.
Smart Images
Figure CN121661740A_ABST
Abstract
Description
[0001] This application is a divisional application of Chinese patent application No. 202411629863.5, filed on November 14, 2024, entitled “Multi-regional two-stage method and electronic device for identifying genuine and counterfeit RMB”, the entire contents of which are incorporated herein by reference. Technical Field
[0002] This invention relates to the field of banknote authentication technology, and in particular to a multi-regional, two-stage method and electronic device for identifying genuine and counterfeit RMB banknotes. Background Technology
[0003] Banknote authentication technology is a crucial technology with a wide range of applications, including banks, shopping malls, supermarkets, and ATMs. With technological advancements, banknote anti-counterfeiting technologies are constantly evolving, from simple optical feature recognition to deep learning technology, continuously improving the anti-counterfeiting capabilities of banknotes.
[0004] In recent years, with the rapid development of deep learning technology, more and more banknote anti-counterfeiting technologies have begun to adopt deep learning for anti-counterfeiting detection. For example, the "A Deep Learning-Based Banknote Anti-counterfeiting Detection Method" released on October 28, 2022, is a deep learning-based banknote anti-counterfeiting technology. This technology uses deep learning to normalize the acquired banknote images, extracts multiple feature regions from the normalized images, and then divides the dataset into training and testing sets according to a preset ratio. The training set is input into a feature extraction network to obtain deep feature maps, which are then input into a classification model for classification training to obtain an initial detection model, and finally, an anti-counterfeiting detection model.
[0005] The inventors discovered that while existing anti-counterfeiting detection technologies improve accuracy and efficiency, they require large datasets and computing resources. Furthermore, model training and optimization also consume a significant amount of time and effort. Therefore, existing detection technologies cannot meet actual production needs. Summary of the Invention
[0006] The embodiments of the present invention are intended to solve at least one of the above-mentioned technical problems.
[0007] In a first aspect, embodiments of the present invention provide a multi-region, two-stage method for authenticating genuine and counterfeit RMB banknotes, comprising: acquiring image information of the banknote sample and correcting and registering the image information to obtain image data of the banknote sample; extracting feature values of at least one anti-counterfeiting identification area in the image data and determining whether the feature values are normal; if the feature values are abnormal, inputting the feature values of the anti-counterfeiting identification area and the images of the functional areas of the banknote sample into a classification network model for training to obtain the classification identification result of the banknote sample, as well as the serial number identification result and version identification result of the banknote sample; analyzing the version identification result, the serial number identification result, and the classification identification result of the banknote sample, and determining the authenticity of the banknote sample; the correction and registration of the image information includes: performing image geometric correction on the image of the banknote sample using a calibration dot matrix and perspective calibration methods, wherein the image geometric correction can improve the angle information and size information of the image of the banknote sample. The location and lighting information are consistent; when there is uneven grayscale distribution in the ticket image, the image brightness field correction is performed on the ticket image; according to the registration algorithm of feature point matching, feature points in the ticket image are extracted and matched with feature points in the preset ticket image; the extraction of feature points in the ticket image and matching of feature points with feature points in the preset ticket image includes: extracting feature points in the ticket image using a feature point detection algorithm; matching the feature points in the ticket image with the feature points in the preset ticket image to obtain at least one pair of feature points; using the RANSAC algorithm to calculate the transformation matrix between the feature point pairs, where the transformation matrix includes an affine transformation matrix and a perspective transformation matrix; transforming the ticket image according to the transformation matrix to align the ticket image with the preset ticket image; checking whether the registration result meets the expected effect, if the registration is not ideal, optimization is performed by adjusting parameters or reselecting a reference image.
[0008] Secondly, embodiments of the present invention provide an electronic device comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute any of the above-described multi-regional dual-stage RMB authenticity identification methods of the present invention.
[0009] Thirdly, embodiments of the present invention provide a storage medium storing one or more programs including execution instructions, which can be read and executed by electronic devices (including but not limited to computers, servers, or network devices, etc.) to perform any of the above-described multi-regional dual-stage RMB authenticity identification methods of the present invention.
[0010] Fourthly, embodiments of the present invention also provide a computer program product, the computer program product including a computer program stored on a storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, causing the computer to execute any of the above-mentioned multi-regional dual-stage RMB authenticity identification methods.
[0011] The method in this invention performs feature analysis on extracted banknote sample images to obtain a preliminary result for determining the authenticity of a region. A deep learning model is then used for secondary confirmation, effectively improving the accuracy of the algorithm's authenticity determination. Combining feature analysis results with network model design enables the specific implementation of a banknote authenticity identification algorithm, further enhancing the accuracy of multi-region, two-stage RMB authenticity identification. A registration algorithm based on feature point matching extracts feature points from the image and matches them with feature points in a reference image, thereby achieving image registration. Attached Figure Description
[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0013] Figure 1 This is a flowchart of an embodiment of the multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes according to the present invention;
[0014] Figure 2 A flowchart of another embodiment of the multi-regional two-stage method for identifying genuine and counterfeit RMB banknotes according to the present invention;
[0015] Figure 3 This is a schematic diagram of the classification network model structure of the multi-regional two-stage RMB authenticity identification method of the present invention;
[0016] Figure 4 This is a schematic diagram of the local convolutional layer structure of the classification network model of the multi-regional two-stage RMB authenticity identification method of the present invention.
[0017] Figure 5 A flowchart illustrating the principle of a multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes, as provided in an embodiment of the present invention.
[0018] Figure 6 This is a schematic block diagram of the structure of an electronic device according to an embodiment of the present invention.
[0019] in, Figure 6 The correspondence between the reference numerals and component names in the attached drawings is as follows:
[0020] 610: Processor; 620: Memory; 630: Input device; 640: Output device. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0022] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.
[0023] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, elements, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.
[0024] In this invention, terms such as "module," "device," and "system" refer to relevant entities applied to a computer, such as hardware, combinations of hardware and software, software, or software in execution. More specifically, for example, an element can be, but is not limited to, a process running on a processor, a processor, an object, an executable element, an execution thread, a program, and / or a computer. Furthermore, an application program or script running on a server, and the server itself, can also be an element. One or more elements may be in an execution process and / or thread, and elements may be localized on a single computer and / or distributed across two or more computers, and may be run on various computer-readable media. Elements can also communicate via local and / or remote processes based on signals having one or more data packets, for example, signals from data interacting with another element in a local system, a distributed system, and / or interacting with other systems via signals over a network on the Internet.
[0025] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising" or "including" include not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0026] This invention provides a multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes. This method can be applied to electronic devices. The electronic devices can be computers, servers, or other electronic products, and this invention does not limit the scope of application.
[0027] Please refer to Figure 1 This illustrates a multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes, provided by an embodiment of the present invention.
[0028] like Figure 1 As shown, in step 101, the image information of the ticket sample is acquired and the image information is corrected and registered to obtain the image data of the ticket sample;
[0029] In step 102, feature values of at least one anti-counterfeiting recognition area in the image data are extracted and it is determined whether the feature values are normal.
[0030] In step 103, if the feature value is abnormal, the feature value of the anti-counterfeiting identification area and the image of the functional area of the ticket sample are input into the classification network model for training to obtain the classification identification result of the ticket sample, as well as the serial number identification result and version identification result of the ticket sample;
[0031] In step 104, the version identification result, serial number identification result, and classification identification result of the ticket sample are analyzed to determine the authenticity of the ticket sample.
[0032] In this embodiment, for step 101, various information about the banknote sample is acquired using various camera and sensor devices. The banknote sample is not limited to different types of banknotes, card types, or other valuable instruments. During the sample acquisition process, all banknote versions and denominations are covered as comprehensively as possible. The various information about the banknote sample can include front and back views, perspective views, test images, fluorescent images, thickness information, magnetic signals, and electrical signals. These image information is corrected and registered to obtain image data of the banknote sample. The image data uses a unified viewpoint to ensure data consistency and comparability. To improve the accuracy and robustness of the banknote authenticity identification algorithm, a large amount of banknote image data needs to be collected. Among this, front and back views, perspective views, test images, fluorescent images, thickness information, magnetic signals, and electrical signals are all crucial data. These data sources can include banknote printing plants, printing technology research institutes, various self-service banks, ATMs, supermarket cash registers, and other similar scenarios.
[0033] Next, in step 102, after obtaining the image data of the banknote sample, the feature value of at least one anti-counterfeiting identification area in the image data is extracted, and its normality is determined. Each banknote sample contains multiple anti-counterfeiting identification areas; only the feature value of one of these areas needs to be obtained for judgment. The feature values of each different anti-counterfeiting identification area are different. After extracting the feature value of an anti-counterfeiting identification area, feature analysis is performed on that feature value to obtain a preliminary result for determining the authenticity of the anti-counterfeiting area. This result is not necessarily normal or abnormal; it is only a preliminary result. Feature analysis of the anti-counterfeiting identification area is performed using the mean, variance, distance between the sample histogram and the genuine banknote histogram, and the cosine distance between the sample and existing genuine banknote samples. Combined with existing thresholds and the range of positive sample feature values, a preliminary determination of the authenticity of the area is made. Extracting the feature value of at least one anti-counterfeiting identification area in the image data is based on the characteristics of the banknote sample (paper money), and ROI region extraction is performed on different anti-counterfeiting identification areas. At the same time, ROI region extraction is performed on some specific functional areas (such as version, serial number, etc.) to obtain the corresponding area images. After obtaining the corresponding region image, the image data needs to be preprocessed. The preprocessed data is then normalized and Gaussian smoothed for noise reduction to improve the accuracy and efficiency of subsequent algorithms.
[0034] Then, for step 103, when the feature value of the anti-counterfeiting identification area is determined to be abnormal, the feature value of the anti-counterfeiting identification area and the image of the functional area of the ticket sample need to be input into the classification network model for training. The functional area is input into PaddleOCR and the classification network for training, with the classification network using the cross-entropy loss function. The trained PaddleOCR network can recognize the serial number in the functional area, and the trained classification network can identify the version of the ticket sample. After inputting the feature value of the anti-counterfeiting identification area and the image of the functional area of the ticket sample into the classification network model for training, the serial number recognition result and version recognition result of the ticket sample can be obtained.
[0035] Finally, for step 104, the version identification results, serial number identification results, and classification identification results of the banknote sample are analyzed to determine its authenticity. For example, if multiple areas of the banknote are detected and analyzed, and one area is identified as a fake sample, then the banknote sample is identified as counterfeit at this stage, resulting in a second authenticity determination. Based on the authenticity determination results output by the network model, combined with the preliminary authenticity determination results, version identification results, and serial number identification results obtained from the first stage feature analysis (feature values of the anti-counterfeiting identification area), the final authenticity of the banknote is determined.
[0036] The method in this application performs feature analysis on the extracted banknote sample image to obtain a preliminary result for determining the authenticity of a region. A deep learning model is then used for secondary confirmation, effectively improving the accuracy of the algorithm's authenticity determination. By combining the feature analysis results with the network model design, a specific implementation of the banknote authenticity identification algorithm is achieved, further enhancing the accuracy of multi-region, two-stage RMB authenticity identification.
[0037] Please refer to Figure 2 This illustrates another multi-regional, two-stage method for authenticating genuine and counterfeit RMB banknotes, provided by an embodiment of the present invention. The flowchart mainly describes the process. Figure 1 The flowchart further defines the step of "correcting and registering the image information" in step 101.
[0038] like Figure 2 As shown, in step 201, the image of the ticket sample is geometrically corrected using calibration dot matrix and perspective calibration methods. The geometrical correction of the image can make the angle information, size information, position information and illumination information of the ticket sample image consistent.
[0039] In step 202, when there is uneven grayscale distribution in the image of the ticket sample, image brightfield correction is performed on the image of the ticket sample;
[0040] In step 203, according to the registration algorithm of feature point matching, feature points in the image of the ticket sample are extracted and matched with feature points in the preset ticket sample image.
[0041] In this embodiment, for step 201, the dot matrix calibration and perspective calibration methods are used to perform geometric correction on the image, eliminating perspective distortion, radial distortion, and tangential distortion caused by the camera during the banknote shooting process. Since there may be angle deviations and distortions during the acquisition process, image correction is necessary to ensure consistency in the angle, size, position, and lighting information of the banknote image. Geometric correction mainly includes: 1. Determining the area to be corrected in the image: Typically, the area to be corrected needs to be selected in the image, such as text, tables, or images. Selecting the correct area is crucial for the effectiveness of perspective correction. 2. Extracting quadrilaterals from the area: The perspective correction algorithm needs to extract the quadrilaterals in the area to be corrected and then convert them into rectangles. Quadrilateral extraction can use techniques such as edge detection and corner detection. 3. Calculating the perspective transformation matrix: The perspective correction algorithm corrects the image by calculating the perspective transformation matrix. The perspective transformation matrix can be calculated by selecting the coordinates of the four vertices and the matrix of the target plane. Applying the perspective transformation matrix: By applying the perspective transformation matrix to the image, the quadrilaterals can be converted into rectangles. In this process, interpolation techniques can be used to fill pixel values. 4. Check the corrected image: Finally, check if the corrected image meets the expected effect. If the image is still distorted or deformed, it can be optimized by adjusting the coordinates of the four vertices and the perspective transformation matrix.
[0042] Next, in step 202, when there is uneven grayscale distribution in the ticket sample image, brightfield correction needs to be performed on the image. Uneven grayscale distribution refers to situations where the grayscale level in the image is not uniformly distributed due to factors such as the light source not providing uniform illumination, vignetting in the camera lens, impurities in the camera sensor, or nonlinearity of the surface of the target object. For this type of distortion, brightfield correction is required, and the process is calculated based on the following formula:
[0043]
[0044] Median is an operator that calculates the median difference between the bright and dark fields, while the correction coefficient is used to correct the grayscale of the corrected image.
[0045] Finally, for step 203, based on the registration algorithm of feature point matching, feature points are extracted from the image of the banknote sample and matched with feature points in the preset banknote sample image. This process is also known as banknote image registration. The banknote registration step is one of the important steps in data processing. It can make the position, size, rotation and other information of the banknote image consistent, thereby facilitating subsequent processing and analysis.
[0046] The method in this application uses a registration algorithm based on feature point matching. It extracts feature points in the image and matches them with feature points in the reference image to achieve image registration.
[0047] It should be noted that the feature point matching registration algorithm provided in this application can be divided into the following steps:
[0048] a) Feature point extraction: Feature points in the image are extracted by using feature point detection algorithms such as SIFT (Scale-invariant feature transform), SURF (Speeded Up Robust Features), and ORB (Oriented FAST and Rotated BRIEF).
[0049] b) Feature point matching: Match the feature points in the image to be registered with the feature points in the reference image to obtain a set of feature point pairs.
[0050] c) Calculate the transformation matrix: Using algorithms such as RANSAC (Random Sample Consensus), calculate the transformation matrix between feature point pairs, such as the affine transformation matrix and the perspective transformation matrix.
[0051] d) Apply the transformation matrix: Transform the image to be registered according to the calculated transformation matrix to make the image to be registered as close as possible to the reference image.
[0052] e) Check the registration results: Finally, it is necessary to check whether the registration results meet the expectations. If the registration is not ideal, it can be optimized by adjusting parameters or reselecting a reference image.
[0053] In some optional embodiments, extracting feature values of at least one anti-counterfeiting identification area from the image data and determining whether the feature values are normal involves performing feature analysis on the anti-counterfeiting identification area of the ticket sample using the mean, variance, distance between the ticket sample histogram and the preset ticket sample histogram, and cosine distance between the ticket sample and the preset ticket sample, and determining the authenticity of the anti-counterfeiting identification area of the ticket sample. A preliminary determination of the authenticity of the area is then made by combining existing thresholds and the feature value range of positive samples.
[0054] In some optional embodiments, images of different versions of tickets and different anti-counterfeiting recognition areas are modeled separately and trained using a classification network model. This classification network model is a partially convolutional model employing a cross-entropy loss function. The classification network model is a network model with a feature value embedding layer. Its core convolutional layer uses partial convolution, resulting in lower latency and higher throughput, effectively reducing computational cost and memory usage while significantly improving the algorithm's efficiency and accuracy. After inputting the image's ROI region and the feature values obtained at the CPU (Central Processing Unit) stage on a GPU (graphics processing unit) device, the network model can obtain the authenticity verification result for that stage.
[0055] The model for further classifying counterfeit currency is the counterfeit currency classification model, and its specific implementation path is as follows:
[0056] The first step is to train a classification network using known types of counterfeit currency;
[0057] The second step is to extract the fully connected (FC) layer of the classification network as the feature vector after dimensionality reduction.
[0058] The third step is to cluster each type of feature vector and calculate the average cluster radius and cluster center of the known types.
[0059] The fourth step is to calculate the distance between the feature to be detected and each class based on the feature vector of the feature to be detected;
[0060] Fifth, if the feature vector of the feature to be detected is more than 5 times the cluster radius from each class, the user is prompted whether to add a new class.
[0061] Step 6: When the user confirms that a new category needs to be added, a new category is added and the classification network is retrained;
[0062] Step 7: When the user determines that no new categories are needed, the features to be detected are added to the known categories and the classification network is retrained.
[0063] This method effectively identifies counterfeit currency by extracting feature vectors and performing cluster analysis. It can also be dynamically updated as new types of counterfeit currency emerge, thus improving the accuracy and robustness of detection.
[0064] It should be noted that the method in this application utilizes a version recognition model, a serial number recognition model, a feature localization model, and an anti-counterfeiting classification model. When a banknote to be inspected passes through the machine, a set of images is acquired. First, this set of images is fed into the version recognition model to identify the version of the banknote. After identifying the version, the feature detection model for that version is read. The feature detection model decomposes the set of images into several features and identifies the type corresponding to each feature. Feature images are then extracted from the set of images. If the extracted feature image is a serial number feature, it is fed into the serial number recognition model to identify the serial number of the banknote. If the extracted feature is an anti-counterfeiting feature, the image undergoes anti-counterfeiting identification to determine whether the feature is genuine or counterfeit. If the feature is identified as counterfeit, the counterfeit banknote type is further identified using the counterfeit banknote classification model.
[0065] In some optional embodiments, the anti-counterfeiting identification area is detected and the results are analyzed. If a counterfeit banknote sample is found in the anti-counterfeiting identification area, the sample is identified as counterfeit at this stage, resulting in a second authenticity determination. The authenticity of the banknote sample is determined based on the feature values of the anti-counterfeiting identification area and the version and serial number identification results output by the classification network model. For example, multiple areas of the banknote are detected and analyzed. If one area is identified as a counterfeit sample, that sample banknote is identified as counterfeit at this stage, resulting in a second authenticity determination. Based on the authenticity determination results output by the network model, combined with the preliminary authenticity determination results, version identification results, and serial number identification results obtained from the first stage feature analysis, the final authenticity of the banknote is determined.
[0066] It should be noted that the technical solution of this application combines feature analysis and deep learning classification algorithms, which can effectively solve the problems of insufficient accuracy of identification results, insufficient robustness in multiple scenarios, and inaccuracy of overall banknote authenticity identification based on single-region feature analysis algorithms in banknote authenticity identification algorithms. Furthermore, the technical solution of this application employs multiple sensors and cameras to acquire various information about banknotes, and the model aggregates multimodal inputs, which can effectively improve the accuracy and recall rate of banknote authenticity identification.
[0067] It should be noted that the technical solution of this application utilizes various information from the front, back, perspective, test images, fluorescence images, thickness information, magnetic signals, and electrical signals of banknotes to analyze and identify them from multiple angles, thereby improving the accuracy and robustness of the algorithm. This application combines feature analysis and deep learning, leveraging both the knowledge and experience of human experts and the advantages of deep learning to improve the accuracy and efficiency of the algorithm. This application also designs a network model based on local convolution, which has lower latency and higher throughput, significantly improving the efficiency and accuracy of the algorithm. The method involved in this application considers the real-time nature and scalability of the algorithm, and can be applied to various scenarios such as self-service banks, ATMs, and supermarket cash registers, meeting the needs of different scenarios.
[0068] It should be noted that, based on extensive experimental data, the multi-regional, two-stage RMB counterfeit identification method provided in this application demonstrates high accuracy and recall, effectively distinguishing between genuine and counterfeit banknotes and ensuring their security and reliability. Furthermore, the algorithm exhibits good real-time performance and scalability, making it applicable to various scenarios such as self-service banks, ATMs, and supermarket cash registers.
[0069] Please refer to this again. Figure 3 , Figure 4 The diagram illustrates a classification network model structure of a multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes, provided by an embodiment of the present invention.
[0070] Network structure diagram as follows Figure 3 As shown, the model uses multiple P modules to efficiently extract features from the input image and reduces the feature resolution (width and height) to 1 / 32 of the input image. Then, global pooling and 1×1 convolutional layers are designed to compress the features into flat, one-dimensional features. The P module consists of a local convolutional layer (PConv), a 1x1 convolutional layer (Conv), a BatchNormal and ReLU activation layer, and a 1x1 convolutional layer connected in series. It also has residual skip connections to prevent excessively deep network structures from degrading the model's capabilities. The P modules include c1×h / 4×w / 4, c2×h / 8×w / 8, c3×h / 16×w / 16, and c4×h / 32×w / 32, where c, h, and w are the number of channels, height, and width, respectively. The P module can be seen in... Figure 3 Partial Convolution (PConv) layers, such as... Figure 4 As shown, this convolution only uses C k Each partial filter processes only a portion of the features, while the remaining features retain the same mapping. This design significantly improves the computational speed of the model and also provides anti-degeneracy functionality similar to skip connections. Wherein, C... pHere, k represents the number of channels and k represents the convolutional kernel. The network model also incorporates three feature values generated during the CPU processing stage: the mean, variance, and the cosine distance between the input sample and the real currency sample. By designing a 1x1 convolutional layer as the feature embedding layer, the feature values are mapped into the network's feature space. Then, the features generated by the cascaded image input are fused together, and finally, a fully connected layer is used to generate the final true / false prediction result.
[0071] Please refer to Figure 5 The present invention provides a flowchart illustrating the principle of the multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes.
[0072] like Figure 5 As shown, various cameras and sensors are first used to acquire various information about the banknote, such as front view, back view, perspective view, test image, fluorescence image, thickness information, magnetic signal, and electrical signal. Then, the image information is corrected and registered to obtain a unified viewpoint. Feature extraction and feature analysis are performed on different anti-counterfeiting recognition areas on the CPU to obtain a preliminary result for determining the authenticity of the region. Simultaneously, feature analysis is performed on some specific functional areas to obtain the banknote's version recognition result and serial number recognition result. Furthermore, this invention designs a network model with a feature value embedding layer, and the basic convolutional layer of this network adopts partial convolution, which has lower latency and higher throughput. On a GPU device, after inputting the image ROI region and the feature values obtained in the previous stage, the network model can obtain the authenticity determination result for this stage. Finally, based on the combined prediction results of the two stages, the authenticity of the banknote is determined.
[0073] By combining data processing and network model design, a specific implementation of a banknote authenticity verification algorithm is achieved. The specific steps are as follows:
[0074] (1) Collect image data of genuine and counterfeit banknotes through multiple sensors.
[0075] (2) Preprocess the banknote image data, including image correction, image registration, anti-counterfeiting identification area extraction, functional area extraction, normalization, etc.
[0076] (3) By means of the mean, variance, distance between the sample histogram and the genuine currency histogram, and cosine distance between the sample and the existing genuine currency sample, the anti-counterfeiting identification area is analyzed. The authenticity of the area is preliminarily determined by combining the existing threshold and the range of positive sample feature values.
[0077] (4) The functional regions are input into PaddleOCR and the partially convolutional classification network proposed in this patent for training. The classification network uses the cross-entropy loss function. The trained PaddleOCR network can recognize the serial number in the functional regions, and the trained classification network can recognize the version of the sample banknote.
[0078] (5) Model images of different versions and different ROI regions separately, and train them using the classification method based on partial convolution proposed in this patent, also using the cross-entropy loss function. The classification network trained in this step can identify the authenticity of sample banknotes in a certain region.
[0079] (6) Detect and analyze the results of multiple areas of the banknote. If one area is identified as a fake sample, the banknote is identified as counterfeit at this stage, and the second determination of the authenticity of the banknote is obtained.
[0080] (7) Based on the authenticity identification results output by the network model, combined with the preliminary authenticity determination results, version identification results and serial number identification results obtained from the first stage feature analysis, the authenticity of the banknotes is determined.
[0081] It should be noted that the multi-region, two-stage RMB counterfeit identification method provided in this application is based on a secondary judgment algorithm. It performs feature analysis on the extracted features to obtain a preliminary regional counterfeit determination result, and then uses a deep learning model for secondary confirmation, effectively improving the accuracy of the algorithm's counterfeit determination. The feature analysis results are combined with network model design to realize the specific implementation of the banknote counterfeit identification algorithm. Simultaneously, a local convolutional network model based on multimodal input is also provided. This model can handle multimodal inputs, namely feature value inputs and image region inputs, and has lower latency and higher throughput, improving the efficiency and accuracy of the algorithm. The multi-region, two-stage RMB counterfeit identification method provided in this application effectively improves the accuracy of the algorithm's counterfeit determination by identifying multiple anti-counterfeiting recognition areas.
[0082] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of combined actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps can be performed in other orders or simultaneously according to the present invention. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention. In the above embodiments, the descriptions of each embodiment have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0083] In some embodiments, the present invention provides a storage medium storing one or more programs including execution instructions, which can be read and executed by electronic devices (including but not limited to computers, servers, or network devices) to perform any of the above-described multi-regional two-stage RMB authenticity identification methods of the present invention.
[0084] In some embodiments, the present invention also provides a computer program product, the computer program product including a computer program stored on a storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to execute any of the above-described multi-regional two-stage RMB authenticity identification methods.
[0085] In some embodiments, the present invention also provides an electronic device comprising: at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform a multi-regional two-stage method for identifying genuine and counterfeit RMB.
[0086] Figure 6 This is a schematic diagram of the hardware structure of an electronic device that performs a multi-regional, two-stage method for authenticating genuine and counterfeit RMB, as provided in another embodiment of this application. Figure 6 As shown, the device includes:
[0087] One or more processors 610 and memory 620, Figure 6 Take the 610 processor as an example.
[0088] The device for performing the multi-regional, two-stage method for authenticating genuine and counterfeit RMB may also include: an input device 630 and an output device 640.
[0089] The processor 610, memory 620, input device 630, and output device 640 can be connected via a bus or other means. Figure 6 Taking the example of a connection between China and Israel via a bus.
[0090] The memory 620, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the multi-regional two-stage RMB authenticity verification method in this embodiment. The processor 610 executes various server functions and data processing by running the non-volatile software programs, instructions, and modules stored in the memory 620, thereby implementing the multi-regional two-stage RMB authenticity verification method described in the above embodiment.
[0091] The memory 620 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 620 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 620 may optionally include memory remotely located relative to the processor 610, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0092] Input device 630 can receive input digital or character information and generate signals related to user settings and function control of the electronic device. Output device 640 may include display devices such as a display screen.
[0093] The one or more modules are stored in the memory 620, and when executed by the one or more processors 610, they execute the multi-regional two-stage RMB authenticity identification method in any of the above method embodiments.
[0094] The above-described product can perform the methods provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects for performing the methods. Technical details not described in detail in this embodiment can be found in the methods provided in the embodiments of this application.
[0095] The electronic devices in this application embodiments exist in various forms, including but not limited to:
[0096] (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and primarily aim to provide voice and data communication. These terminals include smartphones, multimedia phones, feature phones, and low-end phones.
[0097] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, have computing and processing functions, and generally also have mobile internet access capabilities. These terminals include: PDA (personal digital assistant), MID (Mobile Internet Device), and UMPC (Ultra-mobile Personal Computer) devices, etc.
[0098] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes audio and video players, handheld game consoles, e-book readers, as well as smart toys and portable car navigation devices.
[0099] (4) Other airborne electronic devices with data interaction capabilities, such as vehicle-mounted systems installed on vehicles.
[0100] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.
[0101] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A multi-regional, two-stage method for authenticating counterfeit RMB banknotes, characterized in that, include: The image information of the ticket sample is acquired and the image information is corrected and registered to obtain the image data of the ticket sample; Extract feature values from at least one anti-counterfeiting recognition area in the image data and determine whether the feature values are normal; If the feature value is abnormal, the feature value of the anti-counterfeiting recognition area and the image of the functional area of the ticket sample are input into the classification network model for training to obtain the classification recognition result of the ticket sample, as well as the serial number recognition result and version recognition result of the ticket sample; The version identification result, serial number identification result, and classification identification result of the ticket sample are analyzed to determine the authenticity of the ticket sample. The correction and registration of the image information includes: The image of the ticket sample is geometrically corrected using calibration dot matrix and perspective calibration methods, wherein the image geometric correction enables the angle information, size information, position information and illumination information of the ticket sample image to be consistent; When there is uneven grayscale distribution in the image of the ticket sample, image brightfield correction is performed on the image of the ticket sample; According to the registration algorithm of feature point matching, feature points are extracted from the image of the ticket sample and matched with feature points in the preset ticket sample image; The step of extracting feature points from the image of the ticket sample and matching the feature points with feature points in a preset ticket sample image includes: Feature points in the image of the ticket sample are extracted using a feature point detection algorithm; The feature points in the image of the ticket sample are matched with the feature points in the preset ticket sample image to obtain at least one pair of feature points; Using the RANSAC algorithm, the transformation matrix between the feature point pairs is calculated, wherein the transformation matrix includes an affine transformation matrix and a perspective transformation matrix; The ticket sample image is transformed according to the transformation matrix to align it with the preset ticket sample image; Check if the registration results meet the expectations. If the registration is not ideal, optimize it by adjusting the parameters or reselecting a reference image.
2. The multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes according to claim 1, characterized in that, The feature point detection algorithms include SIFT, SURF, and ORB.
3. The multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes according to claim 1, characterized in that, The step of extracting feature values from at least one anti-counterfeiting recognition area in the image data and determining whether the feature values are normal includes: The anti-counterfeiting identification area of the ticket sample is analyzed by means, variance, distance between the ticket sample histogram and the preset ticket sample histogram, and cosine distance between the ticket sample and the preset ticket sample, and the authenticity of the anti-counterfeiting identification area of the ticket sample is determined.
4. The multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes according to claim 1, characterized in that, The analysis of the version identification result, serial number identification result, and classification identification result of the ticket sample, and the determination of the authenticity of the ticket sample, includes: The anti-counterfeiting identification area is detected and the results are analyzed. If there is a counterfeit ticket sample in the anti-counterfeiting identification area, the ticket sample is identified as counterfeit currency at this stage, and the second authenticity determination result of the ticket sample is obtained.
5. The multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes according to claim 4, characterized in that, Before detecting and analyzing the anti-counterfeiting identification area, the following steps are included: The classification network is used to train known fake ticket samples. The fully connected (FC) layer of the classification network is extracted as the feature vector after dimensionality reduction. Each type of feature vector is clustered, and the average cluster radius and cluster center of the known types are statistically analyzed. The distance between the feature vector of the feature to be detected and each class is calculated. When the distance between the feature vector of the feature to be detected and each class exceeds the cluster radius of a preset multiple, the user is prompted whether to add a new class. If necessary, add a new class and retrain the classification network; If not, the features to be detected are added to the known categories and the classification network is retrained.
6. The multi-regional, two-stage method for identifying genuine and counterfeit RMB banknotes according to claim 1, characterized in that, The image information of the obtained ticket sample includes: Image information of the ticket sample is acquired based on sensors and camera devices. The image information of the ticket sample includes a front view, a back view, a perspective view, a test image, a fluorescence image, thickness information, magnetic signals, and electrical signals.
7. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the multi-regional dual-stage RMB authenticity identification method according to any one of claims 1 to 6.
8. A storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the steps of the multi-regional, two-stage method for identifying genuine and counterfeit RMB as described in any one of claims 1 to 6.