An intelligent lock identification method, device, equipment and storage medium

Abstract

The application provides a smart lock identification method and device, equipment and a storage medium, the method comprising: generating a candidate matching area of a vein texture image according to the spatial coordinates of corresponding fingerprint features obtained from a calculated fingerprint image; performing sparse coding on the candidate matching area and the fingerprint image, a first sparse feature vector and a second sparse feature vector, and performing feature fusion on both to obtain a fused feature vector; matching the fused feature vector with a user authorized feature vector, and if the matching score exceeds a first matching threshold, performing an unlocking operation. The application collects a fingerprint image and a vein texture image of a user's finger, generates a candidate matching area of a vein image in combination with the spatial coordinates of the fingerprint features, and on this basis, performs sparse coding and feature fusion on the two biological features, realizes the correlation matching of multi-modal features from the structural level, and based on the fusion verification of the fingerprint features and the vein image, can effectively improve the security of the smart lock.

Description

Technical Field

[0001] This application relates to the field of smart lock technology, and in particular to a smart lock identification method, device, equipment and storage medium. Background Technology

[0002] With the rapid development of smart security and IoT technologies, smart locks have gradually replaced traditional mechanical locks, becoming an important terminal device for security protection in homes, offices, and shared spaces. Existing smart locks widely employ biometric identification methods for identity verification, with single-modal identification, primarily fingerprint recognition, being the most common. However, fingerprints are superficial biometric information, easily copied or forged. Once this data is leaked, it poses a threat to users' property security. Summary of the Invention

[0003] This application provides a smart lock identification method, device, equipment, and storage medium to address the problem of insufficient security of single fingerprint recognition in related technologies.

[0004] The first aspect of this application provides a smart lock identification method, the smart lock identification method comprising: Simultaneously acquire fingerprint and vein texture images of the user's finger; Based on the spatial coordinates obtained by calculating the corresponding fingerprint features in the fingerprint image, a candidate matching region for the vein texture image is generated; The candidate matching region is sparsely encoded to obtain a first sparse feature vector corresponding to the vein feature, and the fingerprint image is similarly sparsely encoded to obtain a second sparse feature vector corresponding to the fingerprint feature. The first sparse feature vector and the second sparse feature vector are aligned and fused to obtain a fused feature vector. The fused feature vector is matched with the user-authorized feature vector. If the matching score exceeds the first matching threshold, the unlocking operation is performed.

[0005] Optionally, in a first implementation of the first aspect of this application, the step of generating a candidate matching region for the vein texture image based on the spatial coordinates obtained by calculating the spatial coordinates of the corresponding fingerprint features in the fingerprint image includes: Fingerprint feature point set is extracted based on the key fingerprint features of the fingerprint image, and a fingerprint coordinate system with the center of the fingertip as the origin is established in the fingerprint image; By performing geometric analysis on the fingerprint feature point set, feature correspondence point pairs between the fingerprint image and the vein texture image are determined, and geometric transformation parameters between the fingerprint coordinate system and the vein image coordinate system are calculated based on the feature correspondence point pairs. Construct a mapping matrix between the fingerprint image and the vein texture image based on the geometric transformation parameters; By performing coordinate mapping operations on the mapping relationship matrix, an initial set of vein candidate regions corresponding to the fingerprint feature points is generated; The initial set of candidate regions is filtered based on the vein pixel density, grayscale gradient, and edge connectivity of the vein texture image to determine the candidate matching regions in the vein texture image.

[0006] Optionally, in the second implementation of the first aspect of this application, the step of obtaining a first sparse feature vector corresponding to the vein feature by sparse coding the candidate matching region, and obtaining a second sparse feature vector corresponding to the fingerprint feature by similar sparse coding of the fingerprint image, includes: A vein feature training dictionary is constructed based on the vein image patches of the candidate matching region. The vein feature training dictionary is composed of multiple vein texture atoms, and each vein texture atom corresponds to the texture structure distribution of the vein image under different directions and scales. By performing sparse coding on the vein image block, the sparse coefficient set of the vein features under the vein feature training dictionary is calculated, and the weights corresponding to the non-zero components in the sparse coefficient set are combined to form the first sparse feature vector of the vein features. A fingerprint feature training dictionary is constructed based on the ridge direction, texture frequency, and minutiae neighborhood distribution in the fingerprint image. By sparsely encoding the feature blocks of the fingerprint image, the sparse coefficient set of the fingerprint image is calculated, and the combination result of the significant components in the sparse coefficient set is determined as the second sparse feature vector of the fingerprint feature.

[0007] Optionally, in a third implementation of the first aspect of this application, the step of aligning and fusing the first sparse feature vector and the second sparse feature vector to obtain a fused feature vector includes: Based on the dimensionality distribution of the first sparse feature vector of the vein feature and the second sparse feature vector of the fingerprint feature, a cross-modal feature mapping model is constructed. The first sparse feature vector and the second sparse feature vector are respectively input into the cross-modal feature mapping model to obtain the projected vein feature normalized vector and fingerprint feature normalized vector; Based on the Euclidean distance and cosine similarity relationship between the vein feature normalized vector and the fingerprint feature normalized vector, the relative alignment parameters in the feature space are determined; Based on the relative alignment parameters, attention weight allocation and feature fusion operations are performed on the vein feature normalization vector and the fingerprint feature normalization vector to generate a fused feature vector.

[0008] Optionally, in the fourth implementation of the first aspect of this application, after the step of generating candidate matching regions for the vein texture image by calculating the spatial coordinates of the corresponding fingerprint features in the fingerprint image, the method further includes: The fingerprint image is pre-matched with the user-authorized fingerprint template; When the pre-matching score exceeds the fingerprint recognition threshold, a secondary matching is performed on the candidate matching regions of the vein texture image based on the user-authorized vein template; The decision to perform the unlocking operation is based on the matching result of the second matching.

[0009] Optionally, in the fifth implementation of the first aspect of this application, before the step of performing secondary matching on the candidate matching regions of the vein texture image according to the user-authorized vein template, the method further includes: The user-authorized vein texture image is divided into two-dimensional meshes to obtain multiple vein mesh units; By randomly scrambling the vein grid units, a vein grid scrambling sequence is generated; Based on the vein grid scrambling sequence, the image blocks corresponding to the vein grid units are rearranged to obtain the user-authorized vein template.

[0010] Optionally, in the sixth implementation of the first aspect of this application, the step of performing secondary matching on the candidate matching regions of the vein texture image based on the user-authorized vein template includes: An encrypted index mapping table is generated based on the user-authorized vein template to record the storage order of the vein grid units; The vein texture image is divided into blocks according to a grid partitioning method consistent with the index mapping table to obtain the grid cells of the vein texture image; The grid cells of the vein texture image are locally matched with the template grid cells indicated by the encrypted index mapping table by means of a sliding scan. The matching score is calculated based on the number of matching grid cells, average similarity, and relative positional consistency. The step of determining whether to perform the unlocking operation based on the matching result of the secondary matching includes: When the matching score exceeds the second matching threshold, the user is determined to be an authorized user.

[0011] A second aspect of this application provides an apparatus for identifying a smart lock, the apparatus being used to implement a smart lock identification method. The apparatus includes: The acquisition module is used to simultaneously acquire fingerprint images and vein texture images of the user's finger; The generation module is used to generate candidate matching regions for the vein texture image based on the spatial coordinates obtained by calculating the corresponding fingerprint features in the fingerprint image; The encoding module is used to perform sparse encoding on the candidate matching region to obtain a first sparse feature vector corresponding to the vein feature, and to perform similar sparse encoding on the fingerprint image to obtain a second sparse feature vector corresponding to the fingerprint feature. The fusion module is used to perform feature alignment and fusion on the first sparse feature vector and the second sparse feature vector to obtain a fused feature vector. The matching module is used to match the fused feature vector with the user-authorized feature vector. If the matching score exceeds the first matching threshold, the unlocking operation is performed.

[0012] A third aspect of this application provides an electronic device, including a memory and a processor, wherein the processor is configured to execute a computer program stored in the memory, and when the processor executes the computer program, it implements the steps of the smart lock identification method provided in the first aspect of this application.

[0013] The fourth aspect of this application provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the steps of the smart lock identification method provided in the first aspect of this application.

[0014] In summary, the smart lock identification method, device, equipment, and storage medium provided in this application simultaneously acquire fingerprint images and vein texture images of a user's finger; generate candidate matching regions for the vein texture image based on the spatial coordinates obtained by calculating the corresponding fingerprint features in the fingerprint image; perform sparse coding on the candidate matching regions to obtain a first sparse feature vector corresponding to the vein features, and perform similar sparse coding on the fingerprint image to obtain a second sparse feature vector corresponding to the fingerprint features; perform feature alignment and fusion on the first and second sparse feature vectors to obtain a fused feature vector; match the fused feature vector with the user-authorized feature vector, and if the matching score exceeds a first matching threshold, perform an unlocking operation. This application, by simultaneously acquiring fingerprint images and vein texture images of a user's finger, generating candidate matching regions for the vein image by combining the spatial coordinates of the fingerprint features, and performing sparse coding and feature fusion on the two biometric features, achieves the association and matching of multimodal features at the structural level. Based on the fusion verification of fingerprint features and vein images, it can effectively improve the security of smart locks. Attached Figure Description

[0015] Figure 1 A flowchart illustrating the smart lock identification method provided in this application embodiment; Figure 2 A schematic diagram of the program modules of the smart lock identification method device provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0016] To make the inventive objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] To address the insufficient security of single fingerprint recognition in related technologies, embodiments of this application provide a smart lock identification method, such as... Figure 1 This is a flowchart illustrating the smart lock identification method provided in this embodiment. The smart lock identification method includes the following steps: Step 110: Simultaneously acquire fingerprint images and vein texture images of the user's finger.

[0018] Specifically, the system uses a dual-sensor module (visible light and near-infrared) to simultaneously acquire data about the user's finger. The visible light sensor records the texture information of the fingerprint ridges on the finger surface, while the near-infrared sensor acquires the transmission or reflection images of the veins inside the finger through illumination with a specific wavelength of light. A synchronous trigger signal is used during the dual-channel acquisition process to ensure that the two types of images correspond in time and space, guaranteeing data consistency in subsequent analysis.

[0019] Step 120: Generate candidate matching regions for the vein texture image based on the spatial coordinates obtained by calculating the corresponding fingerprint features in the fingerprint image.

[0020] Specifically, by performing structural analysis on the acquired fingerprint images, the direction of ridges, intersections of ridges, and distribution of minutiae are identified. A spatial coordinate system with the center of the fingertip as the origin is established based on the coordinates of the feature points. This coordinate system is used to describe the geometric positional relationship of fingerprint features in the image. Then, based on this spatial coordinate information, the vein image is mapped to the region corresponding to the fingerprint image, thereby generating candidate matching regions for the vein texture image.

[0021] Step 130: Perform sparse coding on the candidate matching region to obtain the first sparse feature vector corresponding to the vein feature, and perform similar sparse coding on the fingerprint image to obtain the second sparse feature vector corresponding to the fingerprint feature.

[0022] Specifically, sparse coding analysis is performed on candidate matching regions of vein images. Using a trained vein feature dictionary, image patches are represented as combinations of sparse coefficients, and salient features of the vein texture are extracted to form the first sparse feature vector. Simultaneously, the fingerprint image undergoes the same sparse representation process to capture ridge direction and minutiae structural features, generating a second sparse feature vector. This sparse representation method compresses information redundancy and extracts the most discriminative texture patterns in the image, establishing a unified feature representation for subsequent feature fusion.

[0023] Step 140: Align and fuse the first sparse feature vector and the second sparse feature vector to obtain the fused feature vector.

[0024] Specifically, a cross-modal feature mapping model is used to project two sparse feature vectors onto the same feature space. Alignment and fusion algorithms are then used to calculate their spatial correspondence and weight distribution, generating a fused feature vector. This fusion process integrates the surface structure information of fingerprint features with the deep vascular texture information of vein features, enabling the system to possess both internal and external biometric dimensions under a unified vector representation, thus obtaining a more stable feature description.

[0025] Step 150: Match the fused feature vector with the user-authorized feature vector. If the matching score exceeds the first matching threshold, then perform the unlocking operation.

[0026] Specifically, the fused feature vector is input into the matching unit and compared with the user-authorized feature vector stored in the database to calculate similarity. The score is then used to determine whether a preset matching threshold has been reached. When the similarity between the fused feature and the authorized feature exceeds the set standard, the system controls the actuator to complete the unlocking operation of the smart lock.

[0027] In one optional implementation of this embodiment, the step of generating candidate matching regions for a vein texture image based on the spatial coordinates obtained by calculating the corresponding fingerprint features in the fingerprint image includes: extracting a set of fingerprint feature points based on the key fingerprint features of the fingerprint image, and establishing a fingerprint coordinate system with the center of the fingertip as the origin within the fingerprint image; determining the feature correspondence point pairs between the fingerprint image and the vein texture image through geometric analysis of the fingerprint feature point set, and calculating the geometric transformation parameters between the fingerprint coordinate system and the vein image coordinate system based on the feature correspondence point pairs; constructing a mapping relationship matrix between the fingerprint image and the vein texture image based on the geometric transformation parameters; generating an initial set of vein candidate regions corresponding to the fingerprint feature points by performing coordinate mapping operations on the mapping relationship matrix; and filtering the initial set of candidate regions based on the vein pixel density, grayscale gradient, and edge connectivity of the vein texture image to determine the candidate matching regions in the vein texture image.

[0028] In this embodiment, key features are extracted from the acquired fingerprint image to identify unique texture information such as fingerprint ridges, valleys, and minutiae, and these are converted into a feature point set. These feature points not only record the pixel position of each texture point in the image but also contain local ridge direction information and spatial relationships between adjacent feature points. This feature point set can comprehensively characterize the geometric structure of the fingertip texture. Based on this, a fingerprint coordinate system is established with the fingertip center as the origin. The fingertip center can be located through a core point or triangulation point. This coordinate system allows each feature point to be represented by coordinates relative to the fingertip center, enabling unified management of the overall fingerprint layout and orientation information. By performing geometric analysis on the fingerprint feature point set, the potential spatial correspondence between the fingerprint image and the vein texture image can be inferred by calculating the distance, angle, and relative direction relationships between feature points. This process considers not only the matching of individual feature points but also the geometric consistency between feature points, such as whether the relative angles and distances between adjacent points have corresponding patterns in the vein image, thereby determining candidate point pairs between fingerprint feature points and vein texture images. Based on these corresponding point pairs, rotation, scaling, and translation parameters between the fingerprint coordinate system and the vein image coordinate system are calculated through geometric transformations. These parameters establish a spatial mapping relationship between the two images, mapping the position of fingerprint feature points in the coordinate system to the corresponding spatial region in the vein image. After coordinate mapping operations using the mapping relationship matrix, an initial set of vein candidate regions can be generated. These initial regions have spatial positions in the vein image that match the fingerprint feature points, forming potential vein texture regions. In the initial set, low-intensity or sparse regions can be eliminated by analyzing vein pixel density, and blurry or discontinuous regions can be excluded by identifying texture edges through gray-level gradients. At the same time, vein segments with complete and continuous structures are maintained by edge connectivity evaluation, thus obtaining the final vein candidate matching regions. For example, if a candidate region has a sparse pixel distribution and broken edges, the region is excluded, while regions with clear vein patterns, continuous edges, and obvious gray-level gradients are retained. This ensures that the candidate regions closely correspond to the spatial positions of fingerprint feature points and possess complete vein texture information, providing accurate and high-quality input for subsequent sparse feature extraction and cross-modal fusion.

[0029] In an optional implementation of this embodiment, the steps of obtaining a first sparse feature vector corresponding to the vein feature by sparsely encoding the candidate matching region and obtaining a second sparse feature vector corresponding to the fingerprint feature by similarly sparse encoding the fingerprint image include: constructing a vein feature training dictionary based on the vein image blocks of the candidate matching region, wherein the vein feature training dictionary consists of multiple vein texture atoms, each vein texture atom corresponding to the texture structure distribution of the vein image at different directions and scales; calculating the sparse coefficient set of the vein feature under the vein feature training dictionary by sparsely encoding the vein image blocks, and combining the weights corresponding to the non-zero components in the sparse coefficient set to form the first sparse feature vector of the vein feature; constructing a fingerprint feature training dictionary based on the ridge direction, texture frequency, and minutiae neighborhood distribution in the fingerprint image; calculating the sparse coefficient set of the fingerprint image by sparsely encoding the feature blocks of the fingerprint image, and determining the combination result of the significant components in the sparse coefficient set as the second sparse feature vector of the fingerprint feature.

[0030] In this embodiment, for image patches within the vein candidate matching region, a vein feature training dictionary is first constructed. This dictionary consists of multiple vein texture atoms, each representing the texture structure distribution of the vein image at different directions and scales. These texture atoms can capture the local morphological features of the vein texture, including changes in blood vessel direction, branch nodes, and vein spacing, enabling the dictionary to cover possible local patterns in the vein image. When sparsely encoding the vein image patch, the image patch is represented as a linear combination of several atoms in the training dictionary, and the number of non-zero coefficients is minimized to obtain a sparse coefficient set. Each coefficient represents the contribution weight of the corresponding texture atom to the image patch, and the non-zero coefficients concentrate on the significant vein texture features in the image patch. The weights of the non-zero components in the sparse coefficient set are combined to form the first sparse feature vector of the vein features. This feature vector not only preserves the important spatial information of the vein texture but also compresses redundant data, making the feature representation highly discriminative and computationally efficient. For example, in a vein image patch with obvious blood vessel branches, sparse encoding will select texture atoms whose direction is consistent with the blood vessel direction and assign them higher weights, while the weights of atoms that do not match the blood vessel direction tend to zero, thereby highlighting the main structure of the blood vessel. Simultaneously, a fingerprint feature training dictionary is constructed based on the ridge direction, texture frequency, and minutiae neighborhood distribution in the fingerprint image. This dictionary, by covering different local ridge arrangement patterns, spacing variations, and spatial relationships of minutiae, enables fingerprint features to be efficiently represented in a sparse form. When sparsely encoding feature blocks in the fingerprint image, each feature block is represented as a linear combination of atoms from the training dictionary, and the coefficients are optimized so that only a few salient atoms participate in the combination, forming a sparse coefficient set. The salient components reflect the main texture features of the fingerprint image's ridges and minutiae. Combining these salient components yields the second sparse feature vector of the fingerprint features. This vector highly condenses the core information of the fingerprint image, allowing subsequent cross-modal fusion to directly utilize the most discriminative texture features for comparison. For example, for a fingertip with complex textures, sparse encoding selects atoms in the dictionary whose ridge direction matches the actual ridge and whose minutiae positions match as non-zero coefficients, thus ensuring that the sparse feature vector accurately reflects the unique structure of the fingerprint. This sparse representation method of veins and fingerprints achieves both data compression and feature enhancement, while ensuring the complementarity and discriminative power of information during cross-modal fusion, providing a high-precision and high-efficiency feature foundation for smart lock recognition.

[0031] In an optional implementation of this embodiment, the step of aligning and fusing the first sparse feature vector and the second sparse feature vector to obtain a fused feature vector includes: constructing a cross-modal feature mapping model based on the dimensionality distribution of the first sparse feature vector of the vein feature and the second sparse feature vector of the fingerprint feature; inputting the first sparse feature vector and the second sparse feature vector into the cross-modal feature mapping model respectively to obtain the projected standardized vector of the vein feature and the standardized vector of the fingerprint feature; determining the relative alignment parameters in the feature space based on the Euclidean distance and cosine similarity relationship between the standardized vector of the vein feature and the standardized vector of the fingerprint feature; and performing attention weight allocation and feature fusion operation on the standardized vector of the vein feature and the standardized vector of the fingerprint feature based on the relative alignment parameters to generate a fused feature vector.

[0032] In this embodiment, a cross-modal feature mapping model is constructed based on the dimensional distribution of the first sparse feature vector of vein features and the second sparse feature vector of fingerprint features. This model is designed to map sparse feature vectors from different biological modalities to a unified feature space, making the two types of features spatially comparable and enabling direct similarity measurement. When constructing the mapping model, it is necessary to analyze the dimensional distribution characteristics of the vein and fingerprint feature vectors, such as the response intensity of texture atoms corresponding to each dimension of the vector at different scales and directions, and the distribution density of ridge and minutiae features. This determines the structure of the projection matrix or transformation function, ensuring that the main features of the vector remain intact while redundant information is compressed after mapping. After inputting the first and second sparse feature vectors into the cross-modal feature mapping model, projected standardized vectors of vein and fingerprint features are obtained. These standardized vectors are normalized in a unified feature space, ensuring that the length and scale of each vector are consistent, thereby eliminating differences in the numerical range and direction of the original features and improving the stability of cross-modal comparison. Subsequently, by calculating the Euclidean distance and cosine similarity between the standardized vectors of vein features and fingerprint features, the relative alignment parameters in the feature space are determined. Euclidean distance reflects the absolute difference in spatial position between the two vectors, while cosine similarity measures the consistency of vector directions. By combining these two metrics, the optimal matching direction and offset of the two modalities in space can be determined. For example, if the Euclidean distance is small and the cosine similarity is low, it indicates that the feature vectors are similar in length but have directional deviations; the relative alignment parameters will adjust the vector directions to enhance consistency. Conversely, if the Euclidean distance is large, it indicates that the two modalities are offset in spatial position; the relative alignment parameters will perform translation or scaling corrections. After obtaining the relative alignment parameters, attention weight allocation and feature fusion operations are performed on the standardized vectors of vein features and fingerprint features. The attention weight allocation is adjusted according to the reliability and discriminative power of each feature in the current environment. For example, when the fingerprint image has high clarity but the vein image has insufficient illumination, the weight of the fingerprint feature is increased to improve the fusion effect, while when the vein texture is complete and clear, the weight of the vein feature is correspondingly enhanced. The fusion operation linearly or nonlinearly combines the weighted feature vectors of the two types to generate a fused feature vector. This vector retains the deep vein information of the vein texture while incorporating the ridge and minutiae features of the fingerprint. This results in a final vector with high discriminative power and robustness in distinguishing different users, and it efficiently reflects the comprehensive information of the user's biometrics in subsequent matching and unlocking decisions. Through this cross-modal alignment and fusion method, the smart lock system can maintain the accuracy and security of recognition under different environments and acquisition conditions, achieving effective complementarity between fingerprint and vein features.

[0033] In one optional embodiment of this example, after the step of generating candidate matching regions for the vein texture image by calculating the spatial coordinates of the corresponding fingerprint features in the fingerprint image, the method further includes: pre-matching the fingerprint image with a user-authorized fingerprint template; when the pre-matching score exceeds the fingerprint recognition threshold, performing a secondary match on the candidate matching regions of the vein texture image based on the user-authorized vein template; and determining whether to perform an unlocking operation based on the matching result of the secondary match.

[0034] In this embodiment, after generating candidate matching regions for vein texture images by calculating the spatial coordinates of corresponding fingerprint features in the fingerprint image, the acquired fingerprint image is first pre-matched with the user-authorized fingerprint template. Pre-matching is achieved by comparing the similarity between the feature point set in the fingerprint image and the template feature point set. The matching algorithm calculates the positional deviation between feature points, the difference in ridge direction, and the correspondence of minutiae, thereby obtaining a fingerprint matching score. This score reflects the degree of structural and textural similarity between the acquired fingerprint and the authorized template. If the pre-matching score exceeds a set fingerprint recognition threshold, it indicates that the acquired fingerprint is highly consistent with the authorized user, and the system enters a secondary matching stage based on vein features. In the secondary matching, the system uses the user-authorized vein template to accurately compare the previously generated candidate vein regions. The vein template contains the spatial layout and vein direction of the blood vessel texture under the user's fingertip. By locally aligning the candidate region with the vein template in image space, the texture consistency within the matching region can be calculated, including the direction matching of vein direction, the consistency of vein spacing, and the connectivity of the skeleton structure, thereby generating a vein matching score. During the matching process, by performing local comparisons of multiple grid units in the candidate region and combining the matching results of each grid unit, the precision and robustness of the matching can be enhanced. For example, even under uneven fingertip pressure or changing lighting conditions, the system can still identify areas with a height consistent with the template. Based on the secondary matching results, the system makes a comprehensive judgment on the fused matching score. When the vein matching score reaches or exceeds a preset threshold, combined with the previous fingerprint pre-matching score, the system determines that the user is an authorized user and performs the unlocking operation; otherwise, the smart lock remains locked. Through this dual verification mechanism of fingerprint pre-matching and vein secondary matching, erroneous or forged fingerprint inputs can be effectively filtered out. At the same time, the internal structure of the vein texture provides additional security, thereby ensuring the accuracy of recognition and resistance to attacks under different environmental conditions. For example, if there is slight dirt on the user's fingertip surface, the fingerprint pre-matching score may be slightly lower, but if the vein texture is complete and the matching degree is high, the secondary matching can still provide sufficient judgment basis to ensure that legitimate users can unlock successfully, while illegal inputs cannot pass verification due to vein texture mismatch, thus achieving high security and reliability of smart lock recognition.

[0035] It should be noted that before the smart lock performs matching, the fingerprint and vein images are evaluated for feature quality separately. Fingerprint feature quality can be scored using indicators such as ridge clarity, number of minutiae, and contrast, while vein feature quality can be scored using indicators such as skeleton integrity, grayscale gradient strength, and vessel connectivity. After obtaining the quality scores for the two modalities, they are standardized to values ​​between 0 and 1 to form fingerprint quality weights and vein quality weights. Then, pre-matching is performed. The fingerprint image is pre-matched with the user-authorized template, and the matching score is calculated. Before using the fused feature vector, a fusion judgment threshold is defined. This threshold can be set based on historical matching data and system security policies. For example, setting the fusion judgment threshold to 0.7 means that when the fingerprint matching score is higher than 0.7, the fingerprint information is reliable, and the fused feature vector can be considered. The judgment logic is as follows: if the matching score is ≥ the fusion judgment threshold and both the fingerprint quality weight and the vein quality weight are high (exceeding the set quality threshold), it indicates that both the fingerprint and vein image quality are reliable. In this case, the fused feature vector can be used for cross-modal matching because the fused feature provides the highest discriminative power under the complementary information of the two modalities. If the matching score is greater than or equal to the fusion judgment threshold but the vein image quality is low, fusion matching is skipped, and the fingerprint matching result is relied upon directly, or a secondary vein matching is performed to increase security. If the matching score is less than the fusion judgment threshold and the vein quality is high, fusion features are ignored, and secondary vein matching is used directly as the main judgment criterion to prevent low-quality fingerprints from affecting the accuracy of fusion matching. If both modalities are of low quality or fingerprint pre-matching fails, the matching request is directly rejected, and the lock remains locked. In actual systems, dynamic weights and thresholds can be set for fusion feature matching and secondary matching respectively, so that the judgment process can adaptively adjust according to environmental factors such as lighting, finger pressure, and humidity. For example, if unstable finger contact causes the fingerprint image to be blurry, but the vein image is complete and clear, the system will automatically increase the priority of secondary vein matching, while when both modalities are clear, fusion feature matching takes priority to balance accuracy and efficiency. Through this dynamic judgment mechanism, the optimal matching method can be intelligently selected, enabling the system to improve response speed while ensuring high security, and to be robust to environmental fluctuations.

[0036] In an optional implementation of this embodiment, before the step of performing secondary matching on the candidate matching regions of the vein texture image based on the user-authorized vein template, the method further includes: dividing the user-authorized vein texture image into a two-dimensional grid to obtain multiple vein grid units; generating a vein grid scrambling sequence by randomly shuffling the vein grid units; and rearranging the image blocks corresponding to the vein grid units according to the vein grid scrambling sequence to obtain the user-authorized vein template.

[0037] In this embodiment, to enhance the security of the vein template in the smart lock recognition system, the user-authorized vein texture image needs to be divided into a two-dimensional mesh. The entire vein image is divided into several vein mesh units of uniform size. Each mesh unit contains local vascular texture information, including vein direction, branch nodes, and grayscale gradient changes. This division decomposes the overall vein texture into local structures, allowing each mesh unit to independently carry some biometric information, thus facilitating subsequent encryption processing. When the mesh units are randomly shuffled, a random permutation sequence is generated to rearrange the position order of each mesh unit in the original image. The shuffled sequence is called the vein mesh scrambling sequence. This sequence records the position index of each mesh unit in the original image, ensuring that the integrity of the image block is maintained during the rearrangement process, while achieving unpredictability in position. For example, if a vein image is divided into a 6×6 grid, with each cell numbered from 1 to 36, the scrambling sequence could be in the order of 12, 3, 25, 7, etc. This means that a grid cell originally in the upper left corner of the image might be repositioned to the middle or lower right corner, thus changing the overall visual layout, but the vein texture within each grid cell remains unchanged. During rearrangement based on the scrambling sequence, the image blocks corresponding to each vein grid cell are recombined according to the positions specified in the scrambling sequence to generate a user-authorized vein template. This template may visually differ significantly from the original vein image, but the original order can be restored internally by using the scrambling sequence, achieving reversible encryption of the template. This method not only increases the vein template's resistance to attacks, making it difficult for external attackers to directly identify the true spatial distribution of the vein texture even if they obtain the image data, but also allows the system to use the scrambling sequence to restore the correct positions of the grid cells in subsequent secondary matching, achieving precise local matching with the acquired vein image. For example, if a grid cell contains a key vascular branch, after perturbation, the cell is moved to another location in the image. While the direction of the blood vessel cannot be directly determined by external observation, the system can still find its original location through sequence mapping during comparison, thus ensuring matching accuracy. The entire process, through two-dimensional grid partitioning, disordered scrambling, and rearrangement according to the perturbation sequence, achieves spatial structural obfuscation of the vein template while maintaining the integrity of local texture information, providing a highly secure biometric storage and matching mechanism for smart locks.

[0038] In this embodiment, the step of performing secondary matching on candidate matching regions of a vein texture image based on a user-authorized vein template includes: generating an encrypted index mapping table that records the storage order of vein grid units based on the user-authorized vein template; dividing the vein texture image into blocks according to a grid division method consistent with the index mapping table to obtain grid units of the vein texture image; performing local matching between the grid units of the vein texture image and the template grid units indicated by the encrypted index mapping table using a sliding scan method; calculating a matching score based on the number of matched grid units, average similarity, and relative position consistency; and determining whether to perform an unlocking operation based on the matching result of the secondary matching, including: determining the user as an authorized user when the matching score exceeds a second matching threshold.

[0039] In this embodiment, to achieve secure matching of vein textures, an encrypted index mapping table recording the storage order of vein grid units needs to be generated based on the user-authorized vein template. This mapping table contains the position number of each grid unit in the template and the encrypted index information. The encryption process can be achieved by performing an XOR operation or permutation encoding on the original index, making it impossible for external parties to directly infer the original position of each grid unit in the template, while ensuring that the system can correctly recover the position of the grid units based on the mapping table during matching. Subsequently, the acquired vein texture image is divided into blocks according to the same grid division method as the index mapping table to obtain the grid units of the vein texture image. These grid units completely correspond to the template grid units in size and arrangement, enabling accurate comparison at the local level in the subsequent matching process. For local matching, each grid unit of the vein texture image is compared with the template grid unit indicated by the encrypted index mapping table using a sliding scan method. The sliding scan allows for slight positional shifts to accommodate slight shifts or rotations caused by the user's finger during the acquisition process, thereby improving the robustness of the matching. During the matching process, the similarity between each grid cell and its corresponding template cell is calculated, taking into account texture direction, grayscale gradient, and skeleton connectivity. The number of successfully matched grid cells is recorded, and the average similarity and relative positional consistency of the matched cells are also calculated. These indicators reflect the degree of matching of the overall vein texture in terms of local structure and global layout. For example, if the blood vessel direction, branch nodes, and grayscale pattern of a certain image grid cell are consistent with the template, the cell is counted as a matched grid cell. If its relative positional deviation is within an acceptable range, the positional consistency score is increased. By weighting the number of matched grid cells, average similarity, and relative positional consistency, a final vein matching score can be generated. This score comprehensively reflects the consistency between the acquired image and the user-authorized template in terms of local texture and global structure. When determining whether to perform the unlocking operation based on the secondary matching results, the matching score is compared with a preset second matching threshold. When the matching score exceeds the threshold, the system determines that the user is an authorized user and performs the unlocking operation; otherwise, the smart lock remains locked. This local matching method based on an encrypted index mapping table can not only handle finger position offsets and slight rotations, but also enhance the security of the vein template through encryption. This makes it impossible to directly infer the original vein grid layout even if image data is obtained externally, thus achieving highly secure and robust user authentication.

[0040] According to the intelligent lock recognition method provided in this application, fingerprint images and vein texture images of a user's finger are collected simultaneously. Candidate matching regions of vein images are generated by combining the spatial coordinates of fingerprint features. On this basis, sparse coding and feature fusion of the two biometric features are performed to achieve the association and matching of multimodal features at the structural level. Based on the fusion verification of fingerprint features and vein images, the security of intelligent locks can be effectively improved.

[0041] Figure 2 This application provides an embodiment of a smart lock identification method and apparatus, which can be used to implement the smart lock identification method in the foregoing embodiments. For example... Figure 2 As shown, the smart lock identification method device mainly includes: The acquisition module 10 is used to simultaneously acquire fingerprint images and vein texture images of the user's finger; The generation module 20 is used to generate candidate matching regions for the vein texture image based on the spatial coordinates obtained by calculating the corresponding fingerprint features in the fingerprint image. The encoding module 30 is used to perform sparse encoding on the candidate matching region to obtain the first sparse feature vector corresponding to the vein feature, and to perform similar sparse encoding on the fingerprint image to obtain the second sparse feature vector corresponding to the fingerprint feature. The fusion module 40 is used to perform feature alignment and fusion on the first sparse feature vector and the second sparse feature vector to obtain a fused feature vector. The matching module 50 is used to match the fused feature vector with the user-authorized feature vector. If the matching score exceeds the first matching threshold, the unlocking operation is performed.

[0042] In an optional implementation of this embodiment, the generation module is specifically used for: extracting a fingerprint feature point set based on the key fingerprint features of the fingerprint image, and establishing a fingerprint coordinate system with the center of the fingertip as the origin within the fingerprint image; determining the feature correspondence point pairs between the fingerprint image and the vein texture image through geometric analysis of the fingerprint feature point set, and calculating the geometric transformation parameters between the fingerprint coordinate system and the vein image coordinate system based on the feature correspondence point pairs; constructing a mapping relationship matrix between the fingerprint image and the vein texture image based on the geometric transformation parameters; generating an initial set of vein candidate regions corresponding to the fingerprint feature points by performing coordinate mapping operations on the mapping relationship matrix; and filtering the initial set of candidate regions based on the vein pixel density, grayscale gradient, and edge connectivity of the vein texture image to determine candidate matching regions in the vein texture image.

[0043] In an optional implementation of this embodiment, the encoding module is specifically used for: constructing a vein feature training dictionary based on vein image blocks of candidate matching regions. The vein feature training dictionary consists of multiple vein texture atoms, each vein texture atom corresponding to the texture structure distribution of the vein image under different directions and scales; calculating the sparse coefficient set of vein features under the vein feature training dictionary by sparsely encoding the vein image blocks, and combining the weights corresponding to the non-zero components in the sparse coefficient set to form the first sparse feature vector of the vein features; constructing a fingerprint feature training dictionary based on the ridge direction, texture frequency, and minutiae neighborhood distribution in the fingerprint image; calculating the sparse coefficient set of the fingerprint image by sparsely encoding the feature blocks of the fingerprint image, and determining the combination result of the significant components in the sparse coefficient set as the second sparse feature vector of the fingerprint features.

[0044] In an optional implementation of this embodiment, the fusion module is specifically used to: construct a cross-modal feature mapping model based on the dimensionality distribution of the first sparse feature vector of the vein feature and the second sparse feature vector of the fingerprint feature; input the first sparse feature vector and the second sparse feature vector into the cross-modal feature mapping model respectively to obtain the projected standardized vector of the vein feature and the standardized vector of the fingerprint feature; determine the relative alignment parameters in the feature space based on the Euclidean distance and cosine similarity relationship between the standardized vector of the vein feature and the standardized vector of the fingerprint feature; and perform attention weight allocation and feature fusion operation on the standardized vector of the vein feature and the standardized vector of the fingerprint feature based on the relative alignment parameters to generate a fused feature vector.

[0045] In an optional implementation of this embodiment, the matching module is further configured to: pre-match the fingerprint image with the user-authorized fingerprint template; when the pre-match score exceeds the fingerprint recognition threshold, perform secondary matching on the candidate matching region of the vein texture image according to the user-authorized vein template; and determine whether to perform the unlocking operation based on the matching result of the secondary matching.

[0046] In an optional implementation of this embodiment, the generation module is further configured to: divide the user-authorized vein texture image into a two-dimensional grid to obtain multiple vein grid units; generate a vein grid scrambling sequence by randomly shuffling the vein grid units; and rearrange the image blocks corresponding to the vein grid units according to the vein grid scrambling sequence to obtain the user-authorized vein template.

[0047] In an optional implementation of this embodiment, the matching module is specifically used for: generating an encrypted index mapping table that records the storage order of vein grid units according to the user-authorized vein template; dividing the vein texture image into blocks according to a grid division method consistent with the index mapping table to obtain the grid units of the vein texture image; performing local matching between the grid units of the vein texture image and the template grid units indicated by the encrypted index mapping table through a sliding scan method; calculating a matching score based on the number of matched grid units, average similarity, and relative position consistency; and determining whether to perform an unlocking operation based on the matching result of the second matching, including: when the matching score exceeds a second matching threshold, determining the user as an authorized user.

[0048] According to the intelligent lock recognition method and device provided in this application, fingerprint images and vein texture images of a user's finger are acquired simultaneously. Candidate matching regions of vein images are generated by combining the spatial coordinates of fingerprint features. On this basis, sparse coding and feature fusion of the two biometric features are performed to achieve the association and matching of multimodal features at the structural level. Based on the fusion verification of fingerprint features and vein images, the security of intelligent locks can be effectively improved.

[0049] According to the scheme provided in this application Figure 3 An electronic device is provided as an embodiment of this application. This electronic device can be used to implement the smart lock identification method in the foregoing embodiments, and mainly includes: The system includes a memory 301, a processor 302, and a computer program 303 stored on the memory 301 and executable on the processor 302. The memory 301 and the processor 302 are connected via communication. When the processor 302 executes the computer program 303, it implements the smart lock identification method described in the foregoing embodiments. The number of processors can be one or more.

[0050] The memory 301 can be a high-speed random access memory (RAM) or a non-volatile memory, such as a disk storage device. The memory 301 is used to store executable program code, and the processor 302 is coupled to the memory 301.

[0051] Furthermore, embodiments of this application also provide a computer-readable storage medium, which may be disposed in the electronic device described in the above embodiments, and the computer-readable storage medium may be as described above. Figure 3 The memory in the illustrated embodiment.

[0052] The computer-readable storage medium stores a computer program that, when executed by a processor, implements the smart lock identification method described in the foregoing embodiments. Furthermore, the computer-readable storage medium can also be a USB flash drive, external hard drive, read-only memory (ROM), RAM, magnetic disk, or optical disk, or any other medium capable of storing program code.

[0053] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0054] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0055] The above-described 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 smart lock identification method, characterized in that, include: Simultaneously acquire fingerprint and vein texture images of the user's finger; Based on the spatial coordinates obtained by calculating the corresponding fingerprint features in the fingerprint image, a candidate matching region for the vein texture image is generated; The candidate matching region is sparsely encoded to obtain a first sparse feature vector corresponding to the vein feature, and the fingerprint image is similarly sparsely encoded to obtain a second sparse feature vector corresponding to the fingerprint feature. The first sparse feature vector and the second sparse feature vector are aligned and fused to obtain a fused feature vector. The fused feature vector is matched with the user-authorized feature vector. If the matching score exceeds the first matching threshold, the unlocking operation is performed.

2. The intelligent lock identification method according to claim 1, characterized in that, The step of generating candidate matching regions for the vein texture image based on the spatial coordinates obtained by calculating the corresponding fingerprint features in the fingerprint image includes: Fingerprint feature point set is extracted based on the key fingerprint features of the fingerprint image, and a fingerprint coordinate system with the center of the fingertip as the origin is established in the fingerprint image; By performing geometric analysis on the fingerprint feature point set, feature correspondence point pairs between the fingerprint image and the vein texture image are determined, and geometric transformation parameters between the fingerprint coordinate system and the vein image coordinate system are calculated based on the feature correspondence point pairs. Construct a mapping matrix between the fingerprint image and the vein texture image based on the geometric transformation parameters; By performing coordinate mapping operations on the mapping relationship matrix, an initial set of vein candidate regions corresponding to the fingerprint feature points is generated; The initial set of candidate regions is filtered based on the vein pixel density, grayscale gradient, and edge connectivity of the vein texture image to determine the candidate matching regions in the vein texture image.

3. The intelligent lock identification method according to claim 1, characterized in that, The step of obtaining a first sparse feature vector corresponding to the vein feature by sparse coding the candidate matching region, and obtaining a second sparse feature vector corresponding to the fingerprint feature by similar sparse coding of the fingerprint image, includes: A vein feature training dictionary is constructed based on the vein image patches of the candidate matching region. The vein feature training dictionary is composed of multiple vein texture atoms, and each vein texture atom corresponds to the texture structure distribution of the vein image under different directions and scales. By performing sparse coding on the vein image block, the sparse coefficient set of the vein features under the vein feature training dictionary is calculated, and the weights corresponding to the non-zero components in the sparse coefficient set are combined to form the first sparse feature vector of the vein features. A fingerprint feature training dictionary is constructed based on the ridge direction, texture frequency, and minutiae neighborhood distribution in the fingerprint image. By sparsely encoding the feature blocks of the fingerprint image, the sparse coefficient set of the fingerprint image is calculated, and the combination result of the significant components in the sparse coefficient set is determined as the second sparse feature vector of the fingerprint feature.

4. The intelligent lock identification method according to claim 1, characterized in that, The step of aligning and fusing the first sparse feature vector and the second sparse feature vector to obtain a fused feature vector includes: Based on the dimensionality distribution of the first sparse feature vector of the vein feature and the second sparse feature vector of the fingerprint feature, a cross-modal feature mapping model is constructed. The first sparse feature vector and the second sparse feature vector are respectively input into the cross-modal feature mapping model to obtain the projected vein feature normalized vector and fingerprint feature normalized vector; Based on the Euclidean distance and cosine similarity relationship between the vein feature normalized vector and the fingerprint feature normalized vector, the relative alignment parameters in the feature space are determined; Based on the relative alignment parameters, attention weight allocation and feature fusion operations are performed on the vein feature normalization vector and the fingerprint feature normalization vector to generate a fused feature vector.

5. The intelligent lock identification method according to claim 1, characterized in that, After the step of generating candidate matching regions for the vein texture image by calculating the spatial coordinates of the corresponding fingerprint features in the fingerprint image, the method further includes: The fingerprint image is pre-matched with the user-authorized fingerprint template; When the pre-matching score exceeds the fingerprint recognition threshold, a secondary matching is performed on the candidate matching regions of the vein texture image based on the user-authorized vein template; The decision to perform the unlocking operation is based on the matching result of the second matching.

6. The intelligent lock identification method according to claim 5, characterized in that, Before the step of performing secondary matching on the candidate matching regions of the vein texture image based on the user-authorized vein template, the method further includes: The user-authorized vein texture image is divided into two-dimensional meshes to obtain multiple vein mesh units; By randomly scrambling the vein grid units, a vein grid scrambling sequence is generated; Based on the vein grid scrambling sequence, the image blocks corresponding to the vein grid units are rearranged to obtain the user-authorized vein template.

7. The intelligent lock identification method according to claim 6, characterized in that, The step of performing secondary matching on the candidate matching regions of the vein texture image based on the user-authorized vein template includes: An encrypted index mapping table is generated based on the user-authorized vein template to record the storage order of the vein grid units; The vein texture image is divided into blocks according to a grid partitioning method consistent with the index mapping table to obtain the grid cells of the vein texture image; The grid cells of the vein texture image are locally matched with the template grid cells indicated by the encrypted index mapping table by means of a sliding scan. The matching score is calculated based on the number of matching grid cells, average similarity, and relative positional consistency. The step of determining whether to perform the unlocking operation based on the matching result of the secondary matching includes: When the matching score exceeds the second matching threshold, the user is determined to be an authorized user.

8. A smart lock identification method and device, characterized in that, The smart lock identification method apparatus is used to implement the smart lock identification method according to claim 1, and the smart lock identification method apparatus includes: The acquisition module is used to simultaneously acquire fingerprint images and vein texture images of the user's finger; The generation module is used to generate candidate matching regions for the vein texture image based on the spatial coordinates obtained by calculating the corresponding fingerprint features in the fingerprint image; The encoding module is used to perform sparse encoding on the candidate matching region to obtain a first sparse feature vector corresponding to the vein feature, and to perform similar sparse encoding on the fingerprint image to obtain a second sparse feature vector corresponding to the fingerprint feature. The fusion module is used to perform feature alignment and fusion on the first sparse feature vector and the second sparse feature vector to obtain a fused feature vector. The matching module is used to match the fused feature vector with the user-authorized feature vector. If the matching score exceeds the first matching threshold, the unlocking operation is performed.

9. An electronic device, characterized in that, Includes memory and processor, of which: The processor is used to execute computer programs stored in the memory; When the processor executes the computer program, it implements the steps in the smart lock identification method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps in the smart lock identification method according to any one of claims 1 to 7.

Images