Intelligent bullet cabinet control management system and method

Abstract

The invention provides an intelligent bullet cabinet control management system and method, relates to the field of bullet cabinet management, and adopts a data analysis technology based on artificial intelligence to judge whether a user who wants to open a bullet cabinet is an authorized user of the bullet cabinet or not by performing fingerprint and iris analysis on the user who wants to open the bullet cabinet. And opening the cartridge cabinet in response to the user being the cartridge cabinet authorized user. Therefore, through a multi-biological-feature fusion mode, the reliability of identity verification of the gun and ammunition cabinet is greatly improved, so that the safety management of guns and ammunitions is more effectively realized.

Description

Technical Field

[0001] The present application relates to the field of gun and ammunition cabinet management, and more specifically, to an intelligent gun and ammunition cabinet control management system and method. Background Art

[0002] A gun and ammunition cabinet is a cabinet specially used to store guns and ammunition, and is mainly used for unified management of guns and ammunition. Guns and ammunition are highly dangerous. If they fall into the hands of criminals, they may be used for criminal activities and cause great harm to social security. In order to ensure that only authorized legal personnel can access guns and ammunition, and reduce the possibility of illegal acquisition and use of guns and ammunition from the source, gun and ammunition cabinets need to perform strict identity authentication. Traditional gun and ammunition cabinets may only rely on a single authentication method, such as a password or simple biometrics (such as fingerprints). This method has low security and can be easily cracked or copied.

[0003] Therefore, a smart gun and ammunition cabinet control and management solution is needed. Summary of the invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a smart gun and ammunition cabinet control management system and method.

[0005] According to one aspect of the present application, there is provided an intelligent gun and ammunition cabinet control and management system, which includes:

[0006] A biometric information collection module for persons to be verified, used to obtain facial images and fingerprint images of persons to be verified collected by the gun and ammunition cabinet;

[0007] A biometric information analysis module for the person to be verified, used for comprehensively analyzing the face image and fingerprint image of the person to be verified to obtain a biometric information feature map of the person to be verified;

[0008] The gun and ammunition cabinet control module is used to decide whether to open the gun and ammunition cabinet based on the information contained in the biometric information feature map of the person to be verified.

[0009] According to another aspect of the present application, a smart gun and ammunition cabinet control and management method is provided, which includes:

[0010] Obtaining facial images and fingerprint images of persons to be verified collected from the gun and ammunition cabinet;

[0011] Comprehensively analyzing the face image and fingerprint image of the person to be verified to obtain a biometric information feature map of the person to be verified;

[0012] Based on the information contained in the biometric information feature map of the person to be verified, it is decided whether to open the gun and ammunition cabinet.

[0013] Compared with the prior art, the intelligent gun and ammunition cabinet control and management system and method provided by the present application adopts artificial intelligence-based data analysis technology, and determines whether the user who wants to open the gun and ammunition cabinet is an authorized user of the gun and ammunition cabinet by performing fingerprint and iris analysis on the user who wants to open the gun and ammunition cabinet, and opens the gun and ammunition cabinet in response to the user being an authorized user of the gun and ammunition cabinet. In this way, through the fusion of multiple biometric features, the reliability of gun and ammunition cabinet identity authentication is greatly improved, thereby more effectively realizing the safe management of guns and ammunition. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0015] Figure 1 This is a system block diagram of the intelligent gun and ammunition cabinet control and management system according to an embodiment of the present application.

[0016] Figure 2 This is a block diagram of a biometric information analysis module for persons to be verified in an intelligent gun and ammunition cabinet control and management system according to an embodiment of the present application.

[0017] Figure 3 This is a block diagram of a gun and ammunition cabinet control module in an intelligent gun and ammunition cabinet control and management system according to an embodiment of the present application.

[0018] Figure 4 This is a flow chart of the intelligent gun and ammunition cabinet control and management method according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0020] As mentioned in the background technology above, gun and ammunition cabinets, as a professional cabinet specially designed for properly storing guns and ammunition, play a vital role in maintaining social security and stability. Its main service objects include units with special missions and responsibilities, which use gun and ammunition cabinets to carry out standardized and centralized unified management of guns and ammunition.

[0021] Guns and ammunition are inherently highly dangerous. Once they accidentally fall into the hands of criminals, they are very likely to become tools for them to carry out various criminal activities, endangering the security and stability of the entire society.

[0022] In order to effectively curb the potential risks of illegal acquisition and use of firearms and ammunition from the source, and to ensure the safety and legality of the use of firearms and ammunition, gun and ammunition cabinets must implement extremely strict identity authentication mechanisms. Traditional gun and ammunition cabinets often rely on a single means of identity authentication, such as simple password settings or relatively basic biometric technologies, such as fingerprint recognition. However, in today's increasingly advanced technology and the increasingly complex and diverse methods used by criminals, this single identity authentication method poses many security risks. Passwords may be known or cracked by others for various reasons, and simple fingerprint recognition technology can easily be used by criminals to confuse through means such as copying fingerprints, which greatly reduces the security of gun and ammunition cabinets. Therefore, a smart gun and ammunition cabinet control and management solution is expected.

[0023] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also demonstrated a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks has provided new solutions and solutions for the control and management of intelligent gun and ammunition cabinets.

[0024] Figure 1 FIG. 1 is a system block diagram of the intelligent gun and ammunition cabinet control and management system according to an embodiment of the present application. Figure 1 As shown, in the intelligent gun and ammunition cabinet control and management system 100, it includes: a biometric information collection module 110 for persons to be verified, which is used to obtain the face image and fingerprint image of the person to be verified collected by the gun and ammunition cabinet; a biometric information analysis module 120 for persons to be verified, which is used to perform a comprehensive analysis on the face image and fingerprint image of the person to be verified to obtain a biometric information feature map of the person to be verified; and a gun and ammunition cabinet control module 130, which is used to decide whether to open the gun and ammunition cabinet based on the information contained in the biometric information feature map of the person to be verified.

[0025] In the embodiment of the present application, the biometric information collection module 110 of the person to be verified is used to obtain the face image and fingerprint image of the person to be verified collected by the gun and ammunition cabinet. It should be understood that the face image of the person to be verified contains the facial feature information of the person to be verified, such as the shape, position, proportional relationship of the facial features, skin texture, skin color, and facial expression features. More importantly, the face image contains the eyeball information that can be used for iris recognition. The iris texture of the eyeball is a unique biological feature of each person, with a high degree of individual difference. The detailed features of its texture are the basis for subsequent iris feature extraction. The fingerprint image of the person to be verified contains the texture information on the surface of the finger, including the direction, bifurcation, termination, ring type, bow type, spiral type and other texture features of the ridge and valley line. The arrangement and combination of these textures and the detailed differences constitute the uniqueness of the fingerprint, which is the key information for identifying the individual identity. At the same time, the fingerprint image also contains the pressure distribution information on the fingerprint surface, because different pressing strengths and angles may cause the deformation of the fingerprint image, but the core features of the fingerprint are relatively stable and can be processed and feature extracted by corresponding algorithms. That is, by collecting facial images, iris information can be easily obtained to achieve iris recognition, thus providing a strong basis for identity verification. At the same time, fingerprints are also a unique biometric feature with uniqueness and stability. The fingerprint features of different people are almost never the same, even identical twins have differences. Fingerprint recognition can provide another reliable source of information for identity verification. Through the combined verification of multiple biometrics, the security of identity verification is greatly improved, avoiding the possible loopholes of a single identification method. In this way, the security control of the gun and ammunition cabinet can be achieved, providing a reliable guarantee for the security management of the gun and ammunition cabinet, ensuring that only authorized personnel can open the gun and ammunition cabinet, thereby preventing the illegal acquisition and use of firearms and ammunition and ensuring social security.

[0026] In the embodiment of the present application, the biometric information analysis module 120 of the person to be verified is used to perform a comprehensive analysis on the face image and fingerprint image of the person to be verified to obtain a biometric information feature map of the person to be verified. Specifically, Figure 2 FIG. 1 is a block diagram of a biometric information analysis module for a person to be verified in a smart gun and ammunition cabinet control and management system according to an embodiment of the present application. Figure 2 As shown, the biometric information analysis module 120 of the person to be authenticated includes: an iris feature extraction unit 121, used for analyzing the face image of the person to be authenticated to obtain an iris feature map; a fingerprint feature extraction unit 122, used for analyzing the fingerprint image of the person to be authenticated to obtain a fingerprint texture feature map; and a biometric information feature fusion unit 123, used for fusing the iris feature map and the fingerprint texture feature map to obtain the biometric information feature map of the person to be authenticated.

[0027] In the embodiment of the present application, the iris feature extraction unit 121 is used to analyze the face image of the person to be authenticated to obtain an iris feature map. Specifically, in the embodiment of the present application, the iris feature extraction unit is used to: pass the face image of the person to be authenticated through an eye target detection network to obtain an eye region of interest image; pass the eye region of interest image through an iris feature encoder to obtain the iris feature map.

[0028] Accordingly, considering that in the entire face image, other facial information except the iris is redundant for iris recognition, it will increase the computational complexity and complexity of subsequent processing, and may even interfere with the extraction of iris features. In order to improve the accuracy and efficiency of identity authentication, it is necessary to focus on the most representative and unique biometric feature part, namely the iris. Therefore, in this application, the face image of the person to be authenticated needs to be passed through the eyeball target detection network to obtain the eyeball area of ​​interest image. Narrowing the processing range to the eyeball area of ​​interest can reduce the interference of irrelevant factors on the extraction of iris features, making the extracted iris features more accurate and stable, and avoiding the influence of other facial features or environmental factors on the iris recognition results, thereby improving the accuracy of identity authentication. And because only the eyeball area of ​​interest is focused on, the potential attack surface is reduced. If the entire face image is processed, it may be forged or interfered by the attacker using the similarity of other facial parts, but only processing the eyeball area is more targeted, reducing the possibility of forgery and attack. It should be noted that the eyeball target detection network is a computer vision algorithm based on deep learning, which can automatically locate and extract the area where the eyeball is located in the input face image. The network is usually trained with a large amount of face image data, which contains information such as various face postures, expressions, lighting conditions, and facial features.

[0029] It should be understood that the texture information of the iris is subtle between different people and is hidden in the original image data. In order to mine these unique texture features from the original eyeball region of interest image so that it can be used as a basis for distinguishing different individuals, in this application, the eyeball region of interest image needs to be passed through an iris feature encoder to obtain the iris feature map. In particular, the iris feature encoder described in this application is a convolutional neural network model containing two-dimensional convolution kernels of different sizes. Two-dimensional convolution kernels of different sizes can capture features of different scales. Large-size convolution kernels can capture more macroscopic texture features, such as the overall texture direction of the iris, while small-size convolution kernels are better at capturing local detail features, such as subtle lines and texture changes in the iris texture. By combining convolution kernels of different sizes, the features of the iris can be fully extracted, taking into account both the overall texture information and the local detail information, thereby improving the integrity and accuracy of iris feature extraction.

[0030] In the embodiment of the present application, the fingerprint feature extraction unit 122 is used to analyze the fingerprint image of the person to be verified to obtain a fingerprint texture feature map. Specifically, in the embodiment of the present application, the fingerprint feature extraction unit is used to: extract a fingerprint local directional gradient histogram from the fingerprint image of the person to be verified; arrange the fingerprint local directional gradient histogram and the fingerprint image of the person to be verified as a fingerprint multi-channel input image; pass the fingerprint multi-channel input image through a fingerprint feature encoder to obtain the fingerprint texture feature map.

[0031] Accordingly, in order to effectively capture the texture features of fingerprints, such as the key features such as the direction, bifurcation, and convergence of the ridges and valleys of the fingerprint, it is necessary to extract the fingerprint local directional gradient histogram from the fingerprint image of the person to be verified. The fingerprint local directional gradient histogram is a statistical histogram used to describe texture features, which is calculated based on the grayscale value changes of pixels in each local area (usually a small pixel block) in the fingerprint image. Specifically, it calculates the distribution of gradients in various directions by calculating the gradient changes of pixels in different directions in each local area. The gradient represents the intensity and direction of the change in pixel grayscale value. For example, at the junction of the ridges and valleys of the fingerprint, the gradient changes are more obvious. The distribution information of these gradient directions constitutes the fingerprint local directional gradient histogram, which can well reflect the texture structure of the fingerprint of the person to be verified. By extracting the fingerprint local directional gradient histogram, the texture features of the fingerprint can be described more accurately, thereby improving the accuracy of fingerprint recognition. Moreover, in the actual process of collecting fingerprint images, it may be affected by many factors, such as the force and angle of finger pressing, the humidity of the skin, the quality of the fingerprint collection device, etc. These factors may cause noise or a certain degree of deformation in the fingerprint image. The fingerprint local directional gradient histogram is robust to these situations to a certain extent. Even if there are some small deformations or noises in the fingerprint image, the overall statistical characteristics of the gradient direction distribution usually do not change much, and it can still effectively represent the basic texture characteristics of the fingerprint.

[0032] It should be understood that the fingerprint local directional gradient histogram and the original fingerprint image contain different types of fingerprint feature information that are very important. Specifically, the original fingerprint image intuitively shows the overall shape and basic texture of the fingerprint, including macro features such as the shape and position of the ridges and valleys. The fingerprint local directional gradient histogram focuses on describing the changes in pixel grayscale values ​​in different directions in the local area of ​​the fingerprint, that is, the microscopic texture features. In order to use these two types of information at the same time for a more comprehensive fingerprint feature extraction, it is necessary to arrange the fingerprint local directional gradient histogram and the fingerprint image of the person to be verified as a fingerprint multi-channel input image in this application. After being arranged as a multi-channel image, the features used for fingerprint verification are richer.

[0033] Accordingly, considering that the fingerprint multi-channel input image contains rich fingerprint information, this information is still in a primitive or primary state, containing a lot of details and noise. In order to abstract this information into more representative and higher-level features, so as to facilitate better subsequent identity recognition, it is necessary to pass the fingerprint multi-channel input image through a fingerprint feature encoder to obtain the fingerprint texture feature map. In particular, the fingerprint feature encoder described in the present application is a convolutional neural network model including a deep and shallow feature fusion module. In a convolutional neural network, shallow features usually contain more detailed texture information, such as local texture features and edge information of fingerprints, while deep features contain more abstract and semantic information, such as the overall texture pattern and large structural features of fingerprints. By fusing deep and shallow features, information at different levels can be fully utilized to form a more comprehensive and distinguishing fingerprint texture feature map. In addition, deep and shallow feature fusion can also improve the adaptability to fingerprint images of different quality and acquisition conditions. For example, when some information in a fingerprint image is missing or blurred, shallow features may provide local detail information, while deep features can provide more macroscopic structural information. The combination of the two can make the extracted fingerprint texture feature map more robust and able to accurately describe the characteristics of the fingerprint in various situations.

[0034] In the embodiment of the present application, the biometric information feature fusion unit 123 is used to fuse the iris feature map and the fingerprint texture feature map to obtain the biometric information feature map of the person to be verified. It should be understood that iris and fingerprint are two different biometric features, each with unique physiological characteristics and stability. Iris features are based on the texture of the iris of the eye, and fingerprint features are derived from the lines on the surface of the finger. There are certain limitations to single biometric verification. For example, fingerprints may be inaccurate due to injury, wear and tear, or the influence of collection equipment, and iris recognition may also be affected by factors such as light and eye diseases. By fusing these two biometric feature maps from different sources, their advantages can be comprehensively utilized to reduce identity authentication errors caused by the failure of a single biometric feature, thereby greatly improving the reliability of identity authentication.

[0035] In the embodiment of the present application, the gun and ammunition cabinet control module 130 is used to determine whether to open the gun and ammunition cabinet based on the information contained in the biometric information feature map of the person to be verified. Specifically, Figure 3 FIG. 1 is a block diagram of a gun and ammunition cabinet control module in an intelligent gun and ammunition cabinet control and management system according to an embodiment of the present application. Figure 3 As shown, the gun and ammunition cabinet control module 130 includes: a biometric feature optimization unit 131, which is used to perform global correlation optimization on the biometric feature map of the person to be verified based on significant feature extraction to obtain an optimized biometric feature map of the person to be verified; an identity recognition result generation unit 132, which is used to pass the optimized biometric feature map of the person to be verified through an identity recognition classifier to obtain an identity recognition result, and the identity recognition result is used to indicate whether the person to be verified is an authorized user; a gun and ammunition cabinet control response unit 133, which is used to open the gun and ammunition cabinet in response to the person to be verified being an authorized user.

[0036] In an embodiment of the present application, the biometric information feature optimization unit 131 is used to perform global correlation optimization based on significant feature extraction on the biometric information feature map of the person to be verified to obtain an optimized biometric information feature map of the person to be verified. In particular, considering that the biometric information feature map of the person to be verified has a higher dimension, it means that the model needs to process more data. This not only increases the computational complexity, but may also lead to overfitting, that is, the model performs well on the training set but performs poorly on unseen data (such as a test set or actual application scenario). Moreover, different channels in the biometric information feature map of the person to be verified may contain repeated or less correlated features. Such excessive redundant features may make it difficult for the model to focus on those key features that are truly helpful for identity authentication, thereby reducing the accuracy of identity recognition. Based on this, it is necessary to optimize the biometric information feature map of the person to be verified to highlight the key features that play a leading role in identity recognition. However, traditional deep learning feature optimization methods usually focus on adjusting local pixel weights, while ignoring the relationship between the overall spatial context distribution in the feature map, which will cause the model to miss some important global features, which may in turn increase the misrecognition rate. Based on this, in the technical solution of the present application, it is necessary to perform global correlation optimization on the biometric information feature map of the person to be verified based on significant feature extraction to obtain an optimized biometric information feature map of the person to be verified.

[0037] Specifically, in an embodiment of the present application, the biometric information feature optimization unit is used to: perform feature deconstruction along the channel dimension on the biometric information feature graph of the person to be verified to obtain a set of biometric information feature vectors of the person to be verified; perform feature decomposition based on the eigenvalue on each biometric information feature vector of the person to be verified in the set of biometric information feature vectors of the person to be verified to obtain a set of principal component eigencoding vectors of the biometric information of the person to be verified; calculate the spatial distance entropy between each group of corresponding biometric information feature vectors of the person to be verified and the principal component eigencoding vectors of the biometric information of the person to be verified in the set of biometric information feature vectors of the person to be verified to obtain an identity recognition space constraint vector composed of multiple spatial distance entropies; calculate the product between the identity recognition space constraint vector and its transposed vector to obtain an identity recognition space constraint matrix; input the identity recognition space constraint matrix into the S-based The spatial constraint activation unit of the igmoid activation function is used to obtain the identity recognition spatial constraint feature matrix; based on the identity recognition spatial constraint feature matrix, the feature constraint is performed on the biometric information feature map of the person to be verified to obtain the optimized biometric information feature map of the person to be verified.

[0038] More specifically, in an embodiment of the present application, each biometric information feature vector of the person to be verified in the set of biometric information feature vectors of the person to be verified is subjected to eigenvalue-based eigendecomposition to obtain a set of principal component eigencoding vectors of the biometric information of the person to be verified, including: processing the set of biometric information feature vectors of the person to be verified according to the following formula to obtain a set of principal component eigencoding vectors of the biometric information of the person to be verified; wherein the formula is:

[0039]

[0040] Among them, V i represents the i-th biometric information feature vector of the person to be verified in the set of biometric information feature vectors of the person to be verified, PCA(V i ) indicates the value of V i Perform eigenvalue-based eigendecomposition, U i Yes V i The corresponding sequence of the eigenvalue decomposition vectors of the biometric information of the person to be verified, Λ i V i The corresponding diagonal matrix of the biometric information of the person to be verified, U i T For U i The transpose of v i1 、v i2 、v im Vi The first, second and mth eigenvalue decomposition vectors of the sequence of eigenvalue decomposition vectors of the biometric information of the person to be verified, λ i1 , im V i The corresponding eigenvalues ​​of the first and mth positions of the diagonal matrix of the biometric information of the person to be verified, V i ' indicates V i The corresponding principal component eigenvalue encoding vector of the biometric information of the person to be verified.

[0041] More specifically, in an embodiment of the present application, calculating the spatial distance entropy between each corresponding set of the biometric information feature vectors of the person to be verified and the set of the principal component intrinsic coding vectors of the biometric information of the person to be verified to obtain an identity recognition space constraint vector composed of multiple spatial distance entropies, including: processing the set of the biometric information feature vectors of the person to be verified and the set of the principal component intrinsic coding vectors of the biometric information of the person to be verified according to the following formula to obtain the identity recognition space constraint vector; wherein the formula is:

[0042]

[0043] Among them, V i represents the i-th biometric information feature vector of the person to be verified in the set of biometric information feature vectors of the person to be verified, V i ' indicates V i The corresponding principal component eigenvalue encoding vector of the biometric information of the person to be verified, ||·||2 2 is the square of the Euclidean norm of the vector, arccosh is the inverse hyperbolic cosine function, d P (V i ,V i ') indicates V i and V i 'The spatial distance entropy between i Represents the i-th spatial distance entropy in the identity recognition spatial constraint vector.

[0044] In view of the above technical problems, in the technical solution of the present application, the biometric information feature map of the person to be verified is subjected to global correlation optimization based on the extraction of significant features. The process begins with the feature deconstruction of the biometric information feature map of the person to be verified along the channel dimension to obtain a set of biometric information feature vectors of the person to be verified. It should be understood that the core idea of ​​this operation is to independently extract the features of different channels by decomposing the biometric information feature map of the person to be verified with multiple channels (usually three-dimensional tensors: height, width and channels), thereby making the feature capabilities represented by each channel explicit in the form of a one-dimensional feature vector. In this solution, the goal is to separate and abstract the significant features implicit in the high-dimensional biometric information feature map of the person to be verified, and to further adapt it to the requirements of spatial constraint modeling and optimization.

[0045] Next, the eigenvalue-based eigendecomposition is performed on each of the biometric information feature vectors of the person to be verified in the set of the biometric information feature vectors of the person to be verified to obtain a set of principal component eigencoding vectors of the biometric information of the person to be verified. It should be understood that the theoretical basis of this step is derived from principal component analysis, and its mathematical principle is to find the most important direction of the data in the feature space by calculating the covariance matrix of the eigenvector and using the eigendecomposition method. In this way, the most representative and discriminative components of the set of biometric information feature vectors of the person to be verified can be extracted, that is, the principal component eigencoding vectors of the biometric information of the person to be verified. This process not only effectively reduces the feature dimension, but also compresses feature redundancy and avoids the influence of noise, providing a more centralized and concise representation for subsequent distance calculations and spatial constraints.

[0046] Then, the spatial distance entropy between each set of corresponding biometric information feature vectors of the person to be verified and the set of principal component intrinsic coding vectors of the biometric information of the person to be verified is calculated to obtain an identity recognition spatial constraint vector composed of multiple spatial distance entropies. It should be understood that this step combines distance measurement with entropy theory to quantify the spatial consistency and distribution difference between each pair of feature vectors and their principal components. Generally, the calculation of spatial distance entropy uses typical distance measurement methods such as Euclidean distance or cosine similarity to capture the separation or aggregation between different features. All calculation results are summarized as a set of identity recognition spatial constraint vectors, which generally reflect the overall complexity of feature distribution and the degree of feature structure. By introducing entropy, an information theory measurement method, the uniformity of feature distribution and the degree of fit of data projection in the principal component space can be effectively characterized. Especially in high-dimensional feature learning, spatial distance entropy modeling provides a global constraint in a statistical sense for describing feature distribution behavior, so that redundant features can be quantitatively controlled and significant parts are emphasized.

[0047] Next, the product between the identity recognition space constraint vector and its transposed vector is calculated to obtain the identity recognition space constraint matrix. It should be understood that this process is based on the outer product calculation principle in tensor algebra, and the spatial quantization result of a single vector is expanded into a global space constraint framework represented in matrix form. The element values ​​of the identity recognition space constraint matrix represent the association weights between the feature vectors in the feature channel dimension, revealing the global relationship in the entire identity recognition feature space. This global dependency modeling in matrix form not only enables local feature relationships to be captured in a larger range, but also provides basic input for subsequent activation and optimization operations. Through matrix operations, the constraints of atomic feature pairs can be extended to larger local and global spatial dependency modeling, which is especially important for high-dimensional multi-channel feature analysis.

[0048] Next, the identity recognition space constraint matrix is ​​input into the space constraint activation unit based on the Sigmoid activation function to obtain the identity recognition space constraint feature matrix. It should be understood that in order to further standardize the weight distribution in the identity recognition space constraint matrix, this matrix is ​​input into the space constraint activation unit based on the Sigmoid activation function to generate the activated identity recognition space constraint feature matrix. The role of the activation unit is to normalize the matrix element values ​​through a nonlinear function so that its range is limited to between 0 and 1. The nonlinear compression property of the Sigmoid activation function can not only smooth the range of variation of the eigenvalues, but also enhance the difference between the high-weight and low-weight feature significance. In this compression process, the noise component is explicitly suppressed, and the feature information with obvious distribution law or significant spatial correlation is highlighted. Therefore, in addition to achieving excellent regularization effect and characteristic normalization, the spatial constraint activation also has the effect of enhancing the network's ability to learn significant features.

[0049] Finally, based on the identity recognition space constraint feature matrix, the feature constraints are applied to the biometric information feature map of the person to be verified to obtain the optimized biometric information feature map of the person to be verified. This step uses the identity recognition space constraint feature matrix as a weighting factor to control the different spatial information in the biometric information feature map of the person to be verified element by element. Specifically, the weight values ​​of the feature areas with strong significance are further amplified, and the weakly related areas are reduced or even shielded, so that the optimized biometric information feature map of the person to be verified mainly focuses on specific areas with semantic significance. This is to better achieve identity authentication.

[0050] In the embodiment of the present application, the identity recognition result generating unit 132 is used to pass the optimized biometric information feature graph of the person to be verified through the identity recognition classifier to obtain the identity recognition result, and the identity recognition result is used to indicate whether the person to be verified is an authorized user. It should be understood that in order to verify the identity of a person based on the rich biometric information contained in the optimized biometric information feature graph of the person to be verified, it is necessary to input the optimized biometric information feature graph of the person to be verified into the identity recognition classifier for processing. The identity recognition classifier learns a large amount of historical data, including biometric information data of authorized users and unauthorized users for training, to understand the characteristic distribution and boundaries of different user identities. In the training stage, the identity recognition classifier will learn how to distinguish the characteristics of different categories of users, and form a data-based classification rule, which is automatically mined from the data, rather than manually pre-set, so that the identity recognition classifier can accurately classify according to the learned pattern when facing the new optimized biometric information feature graph of the person to be verified. The identity recognition result is the direct basis for the subsequent opening of the gun and ammunition cabinet.

[0051] In the embodiment of the present application, the gun and ammunition cabinet control response unit 133 is used to open the gun and ammunition cabinet in response to the person to be verified being an authorized user. It should be understood that when it is shown that the person to be verified is an authorized user, the system will start the operation of opening the gun and ammunition cabinet to facilitate the legal user to access the guns and ammunition; and when it is displayed as an unauthorized user, the system will refuse to open the gun and ammunition cabinet to prevent the illegal acquisition of guns and ammunition, which is the core link in maintaining the security of the gun and ammunition cabinet. The main function of the gun and ammunition cabinet is to safely store guns and ammunition and ensure that only authorized personnel can access them. Responding to the identity authentication of the authorized user by opening the gun and ammunition cabinet is a direct way to achieve the purpose of gun and ammunition cabinet management. This allows legal users to normally obtain guns and ammunition when needed for legal purposes such as performing tasks and training.

[0052] In summary, the intelligent gun and ammunition cabinet control and management system 100 based on the embodiment of the present application is explained, which adopts artificial intelligence-based data analysis technology, performs fingerprint and iris analysis on the user who wants to open the gun and ammunition cabinet to determine whether the user is an authorized user of the gun and ammunition cabinet, and opens the gun and ammunition cabinet in response to the user being an authorized user of the gun and ammunition cabinet. In this way, through the fusion of multiple biometric features, the reliability of the identity authentication of the gun and ammunition cabinet is greatly improved, thereby more effectively realizing the safe management of firearms and ammunition.

[0053] Figure 4 FIG. 1 is a flow chart of a method for controlling and managing an intelligent gun and ammunition cabinet according to an embodiment of the present application. Figure 4As shown, in the intelligent gun and ammunition cabinet control and management method, it includes: S110, obtaining the face image and fingerprint image of the person to be verified collected by the gun and ammunition cabinet; S120, comprehensively analyzing the face image and fingerprint image of the person to be verified to obtain a biometric information feature map of the person to be verified; S130, deciding whether to open the gun and ammunition cabinet based on the information contained in the biometric information feature map of the person to be verified.

[0054] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned intelligent gun and ammunition cabinet control and management method have been described in detail above. Figures 1 to 3 The description of the intelligent gun and ammunition cabinet control and management system has been introduced in detail, and therefore, its repeated description will be omitted.

[0055] In summary, the intelligent gun and ammunition cabinet control and management method based on the embodiment of the present application is explained, which adopts artificial intelligence-based data analysis technology, performs fingerprint and iris analysis on the user who wants to open the gun and ammunition cabinet to determine whether the user is an authorized user of the gun and ammunition cabinet, and opens the gun and ammunition cabinet in response to the user being an authorized user of the gun and ammunition cabinet. In this way, through the fusion of multiple biometric features, the reliability of gun and ammunition cabinet identity authentication is greatly improved, thereby more effectively realizing the safe management of firearms and ammunition.

Claims

1. An intelligent gun and ammunition cabinet control and management system, characterized in that: include: A biometric information collection module for persons to be verified, used to obtain facial images and fingerprint images of persons to be verified collected by the gun and ammunition cabinet; A biometric information analysis module for the person to be verified, used for comprehensively analyzing the face image and fingerprint image of the person to be verified to obtain a biometric information feature map of the person to be verified; The gun and ammunition cabinet control module is used to decide whether to open the gun and ammunition cabinet based on the information contained in the biometric information feature map of the person to be verified.

2. The intelligent gun and ammunition cabinet control and management system according to claim 1 is characterized in that: The biometric information analysis module of the person to be verified includes: An iris feature extraction unit, used for analyzing the face image of the person to be authenticated to obtain an iris feature map; A fingerprint feature extraction unit, used for analyzing the fingerprint image of the person to be authenticated to obtain a fingerprint texture feature map; The biometric information feature fusion unit is used to fuse the iris feature map and the fingerprint texture feature map to obtain the biometric information feature map of the person to be authenticated.

3. The intelligent gun and ammunition cabinet control and management system according to claim 2 is characterized in that: The iris feature extraction unit is used to: Passing the face image of the person to be verified through an eyeball target detection network to obtain an eyeball region of interest image; The eyeball area of ​​interest image is passed through an iris feature encoder to obtain the iris feature map.

4. The intelligent gun and ammunition cabinet control and management system according to claim 3 is characterized in that: The fingerprint feature extraction unit is used to: Extracting a fingerprint local directional gradient histogram from the fingerprint image of the person to be authenticated; Arranging the fingerprint local directional gradient histogram and the fingerprint image of the person to be authenticated into a fingerprint multi-channel input image; The fingerprint multi-channel input image is passed through a fingerprint feature encoder to obtain the fingerprint texture feature map.

5. The intelligent gun and ammunition cabinet control and management system according to claim 4 is characterized in that: The iris feature encoder is a convolutional neural network model including two-dimensional convolution kernels of different sizes, and the fingerprint feature encoder is a convolutional neural network model including a deep and shallow feature fusion module.

6. The intelligent gun and ammunition cabinet control and management system according to claim 5 is characterized in that: The gun and ammunition cabinet control module comprises: A biometric information feature optimization unit, used for performing global correlation optimization based on significant feature extraction on the biometric information feature graph of the person to be verified to obtain an optimized biometric information feature graph of the person to be verified; An identity recognition result generating unit, used for passing the optimized biometric information feature graph of the person to be verified through an identity recognition classifier to obtain an identity recognition result, wherein the identity recognition result is used to indicate whether the person to be verified is an authorized user; The gun and ammunition cabinet control response unit is used to open the gun and ammunition cabinet in response to the person to be verified as an authorized user.

7. The intelligent gun and ammunition cabinet control and management system according to claim 6 is characterized in that: The bio-information feature optimization unit is used to: Performing feature deconstruction along the channel dimension on the biometric information feature graph of the person to be verified to obtain a set of biometric information feature vectors of the person to be verified; Performing eigenvalue-based eigendecomposition on each of the biometric information feature vectors of the to-be-verified identity person in the set of the biometric information feature vectors of the to-be-verified identity person to obtain a set of principal component eigencoding vectors of the biometric information of the to-be-verified identity person; Calculate the spatial distance entropy between each corresponding set of the biometric information feature vectors of the person to be verified and the principal component eigencode vectors of the biometric information of the person to be verified in the set of the biometric information feature vectors of the person to be verified and the principal component eigencode vectors of the biometric information of the person to be verified to obtain an identity recognition spatial constraint vector composed of multiple spatial distance entropies; Calculating the product between the identity recognition space constraint vector and its transposed vector to obtain an identity recognition space constraint matrix; Inputting the identity recognition space constraint matrix into a space constraint activation unit based on a Sigmoid activation function to obtain an identity recognition space constraint feature matrix; Based on the identity recognition space constraint feature matrix, feature constraints are performed on the biometric information feature map of the person to be verified to obtain the optimized biometric information feature map of the person to be verified.

8. A method for controlling and managing an intelligent gun and ammunition cabinet, characterized in that: include: Obtaining facial images and fingerprint images of persons to be verified collected from the gun and ammunition cabinet; Comprehensively analyzing the face image and fingerprint image of the person to be verified to obtain a biometric information feature map of the person to be verified; Based on the information contained in the biometric information feature map of the person to be verified, it is decided whether to open the gun and ammunition cabinet.

9. The intelligent gun and ammunition cabinet control and management method according to claim 8 is characterized in that: Comprehensively analyzing the face image and fingerprint image of the person to be verified to obtain a biometric information feature map of the person to be verified, including: Analyzing the facial image of the person to be authenticated to obtain an iris feature map; Analyzing the fingerprint image of the person to be authenticated to obtain a fingerprint texture feature map; The iris feature map and the fingerprint texture feature map are fused to obtain the biometric information feature map of the person to be authenticated.

10. The intelligent gun and ammunition cabinet control and management method according to claim 9 is characterized in that: Based on the information contained in the biometric information feature map of the person to be verified, deciding whether to open the gun and ammunition cabinet includes: Performing global correlation optimization based on significant feature extraction on the biometric information feature graph of the person to be verified to obtain an optimized biometric information feature graph of the person to be verified; Passing the optimized biometric information feature graph of the person to be verified through an identity recognition classifier to obtain an identity recognition result, wherein the identity recognition result is used to indicate whether the person to be verified is an authorized user; In response to the person to be verified as an authorized user, the gun and ammunition cabinet is opened.

Images