Palm vein recognition model construction method and device, electronic equipment and storage medium
By adjusting the grayscale interval and pooling of the palm vein diagram, a palm vein recognition model is constructed, which solves the problem of high resource consumption in the existing technology, improves the recognition efficiency and accuracy, and facilitates miniaturization and promotion.
Patent Information
- Application Number
- CN202411202190.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-29
AI Technical Summary
The existing palm vein recognition technology requires a large amount of software resources and hardware resources. The construction of identification models requires massive samples and a long training cycle. Relying on intelligent algorithms leads to difficulties in miniaturization and miniaturization, which affects promotion.
By obtaining multiple palm vein diagrams, greyscale interval adjustment and edge feature extraction are performed, pooling is performed according to the pooling degree parameters, identification codes are generated, and parameters are adjusted through the identification generation model until the deviation is less than the threshold, and a palm vein identification model is constructed.
The resource requirements for the identification model are reduced, the recognition efficiency and accuracy are improved, and the resource consumption in early investment and later use is reduced, making it easier to miniaturize and promote.
Smart Images

Figure CN118968573A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of palm vein data recognition technology, and in particular to a palm vein recognition model construction method, device, electronic equipment and storage medium. Background Art
[0002] Palm vein recognition technology is a new generation of biometric identification technology, which identifies the hidden biological characteristics of the human body - vein information. Palm vein recognition technology uses the characteristics of hemoglobin in human blood to absorb near-infrared light to obtain palm vein images, and extracts feature values through positioning, enhancement, feature extraction and other algorithms to verify personal identity. It can perfectly replace codes, cards, coins and certificates.
[0003] Palm vein recognition technology has many unique advantages. First, the vein pattern is the network structure of blood vessels under the human skin, which is almost invisible to the naked eye under natural lighting conditions and can only be obtained from the body when infrared lighting is used. This makes it extremely difficult to steal or forge finger vein information, greatly improving the security of the technology. Secondly, palm vein recognition technology is an internal physiological feature that will not wear out and has high stability and long-term availability. In addition, it also has the characteristics of contactless recognition, and the palm can be recognized without contacting the device, which is both hygienic and convenient.
[0004] At present, palm vein recognition technology mostly uses intelligent algorithms (such as artificial neural networks, support vector machines, etc.) to obtain recognition models through training with massive samples. The disadvantages of this recognition technology are that the construction of the recognition model requires a massive number of samples and a long training and effect verification cycle, and the initial investment is large; the intelligent algorithm that the recognition model relies on depends on the operating system, software platform and hardware, and the investment in both software and hardware is large, and miniaturization and micro-micro-production are relatively difficult, which is not conducive to later promotion.
[0005] Based on this, it is necessary to develop and design a palm vein recognition model construction method. Summary of the invention
[0006] The embodiments of the present invention provide a palm vein recognition model construction method, device, electronic device and storage medium, which are used to solve the problem that palm vein recognition in the prior art consumes a lot of software resources or hardware resources.
[0007] In a first aspect, an embodiment of the present invention provides a method for constructing a palm vein recognition model, comprising: Obtain multiple palm vein maps; Performing grayscale interval adjustment and edge feature extraction on the multiple palm vein images to obtain multiple edge feature images, wherein each edge feature image corresponds to a palm vein image; Pooling each edge feature map according to the first pooling degree parameter, and assigning an identification code to a pooled data block obtained by pooling according to the palm corresponding to the palm vein map, wherein the pooled data blocks corresponding to the same palm correspond to the same identification code; A plurality of sub-data blocks are extracted from the pooled data block and input into the identification generation model. According to the deviation between the identification generated by the model and the identification of the pooled data block, the parameters of the identification generation model are adjusted until the deviation is less than a threshold, thereby obtaining a palm vein recognition model.
[0008] In a possible implementation, the grayscale interval adjustment and edge feature extraction of the multiple palm vein images to obtain multiple edge feature images include: For each of the multiple palm vein maps, the following steps are performed respectively: Acquire a reference queue and a plurality of pixel value intervals, wherein each queue element of the reference queue corresponds to a pixel value interval; Arrange multiple pixel values in the palm vein image according to the size of the values to obtain a pixel value queue; Extracting pixel values from the pixel value queue according to the reference queue to construct a plurality of pixel sets, wherein each pixel set corresponds to a queue element of the reference queue; According to the multiple pixel sets and the multiple pixel value intervals, adjusting the grayscale intervals of the pixels in the palm vein image to obtain a grayscale image; The edge of the grayscale image is extracted using a difference method to obtain an edge feature map.
[0009] In a possible implementation, adjusting the grayscale intervals of pixels in the palm vein image according to the multiple pixel sets and the multiple pixel value intervals to obtain a grayscale image includes: According to the first formula, the plurality of pixel sets and the plurality of pixel value intervals, the grayscale intervals of the pixels in the palm vein image are adjusted to obtain a grayscale image, wherein the first formula is:
[0010] In the formula, is the pixel value after grayscale interval adjustment, is the pixel value before grayscale interval adjustment, is the minimum pixel value in the pixel set, is the maximum pixel value in the pixel set, is the upper limit of the pixel value interval, is the lower limit of the pixel value interval; The step of extracting edges from the grayscale image using a differential method to obtain an edge feature map includes: For each pixel in the edge feature map, a minimum pixel value within a radius of 1 pixel is used to replace the pixel value to obtain an edge convexification map; The edge of the grayscale image is extracted according to a second formula, the grayscale image and the edge convexification image, wherein the second formula is:
[0011] In the formula, is the pixel value of the edge feature map, is the pixel value of the edge convex image, is the pixel value of the grayscale image.
[0012] In a possible implementation manner, pooling each edge feature map according to the first pooling degree parameter includes: For each edge feature map, perform the following steps: Obtaining a first pooling degree parameter, wherein the first pooling degree parameter includes: a pooling dimension and an offset; According to the first position, take out the data block of the pooling dimension from the edge feature map as the data block to be processed; Taking the maximum value or average value of multiple data in the data block to be processed as pooling data; Taking the first position as an index, adding the pooled data to a pooled data block; If the traversal of the edge feature map is not completed, the first position is offset according to the offset, and the process jumps to the step of taking out the data block of the pooling dimension from the edge feature map according to the first position as the data block to be processed.
[0013] In a possible implementation, the identification generation model is:
[0014] In the formula, To identify the output, is the total number of rows of intermediate nodes, For the The weight coefficients of the output nodes, For the Line The first column node The weight coefficients, For the Line The output of the middle node of the column, For the input variables, is the total number of columns of intermediate nodes, is a natural constant, is the bias coefficient.
[0015] In a possible implementation, extracting a plurality of sub-data blocks from the pooled data block and inputting them into the identifier generation model includes: For each pooled data block, perform the following steps: Get the segmentation dimension and number of sub-blocks; Splitting the pooled data block into a plurality of sub-data blocks according to the segmentation dimension; Randomly selecting the number of sub-data blocks from the multiple sub-data blocks as multiple sub-data blocks to be processed; The multiple sub-data blocks to be processed are input into the identification generation model.
[0016] In a possible implementation, adjusting the parameters of the identifier generation model according to the deviation between the identifier generated by the model and the identifier of the pooled data block until the deviation is less than a threshold includes: Acquire a plurality of coefficient arrays, wherein each coefficient array includes data corresponding to a plurality of coefficients in the identification generation model; Substituting the plurality of coefficient arrays into the identification generation model respectively to obtain a plurality of identification code indications, wherein each identification code indication corresponds to a coefficient array; Determine a plurality of identification deviations according to a target identification code and the plurality of identification code indications, wherein each identification deviation corresponds to a coefficient array, and the target identification code is an identification code of a pooled data block from which the plurality of sub-data blocks are derived; If there is a coefficient array whose identification deviation is greater than the deviation threshold, three optimal coefficient arrays are selected from the multiple coefficient arrays according to the multiple identification deviations, and the coefficient array greater than the deviation threshold is adjusted according to a third formula and the three optimal coefficient arrays, wherein the third formula is:
[0017] In the formula, For the The adjusted coefficient array data, For the The adjusted coefficient array data, , , as well as They are the first distance, the second distance, the third distance and the fourth distance respectively. , , as well as are the first coefficient, the second coefficient, the third coefficient and the fourth coefficient respectively, , , as well as are the optimal coefficient arrays. data, the suboptimal coefficient array data, the third best coefficient array data and the optimal coefficient array of the current coefficient array for each iteration individual data; If the traversal of all the pooled data blocks is not completed, multiple sub-data blocks are extracted from the pooled data blocks and input into the identification generation model, and the step of determining multiple identification deviations according to the target identification code and the multiple identification code indications is jumped to; If the traversal of all the pooled data blocks is completed and there is an unadjusted coefficient array in the current traversal process, the identification generation model into which the unadjusted coefficient array is substituted is used as the palm vein recognition model; Otherwise, a new round of traversal of the pooled data blocks is enabled, multiple sub-data blocks are extracted from the traversed pooled data blocks and input into the identification generation model, and the process jumps to the step of determining multiple identification deviations according to the target identification code and the multiple identification code indications.
[0018] In a second aspect, an embodiment of the present invention provides a palm vein recognition model construction device, which is used to implement the palm vein recognition model construction method described in the first aspect or any possible implementation of the first aspect, and the palm vein recognition model construction device includes: An image acquisition module, used to acquire multiple palm vein images; An edge extraction module, used for performing grayscale interval adjustment and edge feature extraction on the plurality of palm vein images to obtain a plurality of edge feature images, wherein each edge feature image corresponds to a palm vein image; A pooling module, used for pooling each edge feature map according to a first pooling degree parameter, and assigning an identification code to a pooled data block obtained by pooling according to a palm corresponding to the palm vein map, wherein pooled data blocks corresponding to the same palm correspond to the same identification code; as well as, The model building module is used to extract multiple sub-data blocks from the pooled data block and input them into the identification generation model. According to the deviation between the identification generated by the model and the identification of the pooled data block, the parameters of the identification generation model are adjusted until the deviation is less than a threshold, thereby obtaining a palm vein recognition model.
[0019] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the steps of the method described in the first aspect or any possible implementation method of the first aspect.
[0020] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect or any possible implementation of the first aspect are implemented.
[0021] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: The embodiment of the present invention discloses a palm vein recognition model construction method, which first obtains multiple palm vein maps; then adjusts the grayscale interval and extracts edge features of the multiple palm vein maps to obtain multiple edge feature maps, wherein each edge feature map corresponds to a palm vein map; then pools each edge feature map according to a first pooling degree parameter, and assigns an identification code to a pooled data block obtained by pooling according to the palm corresponding to the palm vein map, wherein the pooled data blocks corresponding to the same palm correspond to the same identification code; finally, extracts multiple sub-data blocks from the pooled data block and inputs them into an identification generation model, and adjusts the parameters of the identification generation model according to the deviation between the identification generated by the model and the identification of the pooled data block until the deviation is less than a threshold value, thereby obtaining a palm vein recognition model. The implementation method of the present invention extracts edge features after grayscale processing of the palm vein map. Compared with the edge features extracted by binarization, the feature content contained is richer, which is convenient for improving the recognition rate; the implementation method of the present invention extracts part of the data in the pooled data block after pooling and sends it to the model, and fits the identification code to the model, so that when the model is applied, only a small part of the palm vein map needs to be obtained to complete the recognition, which improves the recognition efficiency and accuracy. The implementation method of the present invention does not require an intelligent algorithm, and determines the model parameters by generating multiple sub-data blocks through the palm vein map. Compared with traditional intelligent algorithms, it requires fewer samples, has low dependence on software platforms and hardware platforms, consumes less resources in both early modeling and later use, and is easy to use and promote. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0023] Figure 1 is a flow chart of a palm vein recognition model construction method provided by an embodiment of the present invention; Figure 2 It is a principle diagram of the edge convexification image construction process provided by an embodiment of the present invention; Figure 3is a schematic diagram of a pooling process provided by an embodiment of the present invention; Figure 4 It is a functional block diagram of a palm vein recognition model building device provided by an embodiment of the present invention; Figure 5 It is a functional block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0024] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, and methods are omitted so as not to obscure the description of the present invention with unnecessary details.
[0025] In order to make the purpose, technical solutions and advantages of the present invention more clear, a specific implementation method will be described below in conjunction with the accompanying drawings.
[0026] The following is a detailed description of an embodiment of the present invention. This example is implemented based on the technical solution of the present invention, and provides a detailed implementation method and a specific operation process, but the protection scope of the present invention is not limited to the following embodiments.
[0027] Figure 1 This is a flow chart of a palm vein recognition model construction method provided in an embodiment of the present invention.
[0028] like Figure 1 As shown, it shows a flowchart of the implementation method of the palm vein recognition model construction method provided by the embodiment of the present invention, which is described in detail as follows: In step 101, a plurality of palm vein images are acquired.
[0029] In step 102, grayscale interval adjustment and edge feature extraction are performed on the multiple palm vein images to obtain multiple edge feature images, wherein each edge feature image corresponds to a palm vein image.
[0030] In some implementations, step 102 includes: For each of the multiple palm vein maps, the following steps are performed respectively: Acquire a reference queue and a plurality of pixel value intervals, wherein each queue element of the reference queue corresponds to a pixel value interval; Arrange multiple pixel values in the palm vein image according to the size of the values to obtain a pixel value queue; Extracting pixel values from the pixel value queue according to the reference queue to construct a plurality of pixel sets, wherein each pixel set corresponds to a queue element of the reference queue; According to the multiple pixel sets and the multiple pixel value intervals, adjusting the grayscale intervals of the pixels in the palm vein image to obtain a grayscale image; The edge of the grayscale image is extracted using a difference method to obtain an edge feature map.
[0031] In some implementations, adjusting the grayscale intervals of pixels in the palm vein image according to the multiple pixel sets and the multiple pixel value intervals to obtain a grayscale image includes: According to the first formula, the plurality of pixel sets and the plurality of pixel value intervals, the grayscale intervals of the pixels in the palm vein image are adjusted to obtain a grayscale image, wherein the first formula is:
[0032] In the formula, is the pixel value after grayscale interval adjustment, is the pixel value before grayscale interval adjustment, is the minimum pixel value in the pixel set, is the maximum pixel value in the pixel set, is the upper limit of the pixel value interval, is the lower limit of the pixel value interval; The step of extracting edges from the grayscale image using a differential method to obtain an edge feature map includes: For each pixel in the edge feature map, a minimum pixel value within a radius of 1 pixel is used to replace the pixel value to obtain an edge convexification map; The edge of the grayscale image is extracted according to a second formula, the grayscale image and the edge convexification image, wherein the second formula is:
[0033] In the formula, is the pixel value of the edge feature map, is the pixel value of the edge convex image, is the pixel value of the grayscale image.
[0034] For example, as mentioned above, currently, palm vein recognition is usually performed by binary edge processing of the palm vein image and then feeding it into a model constructed by an artificial neural network for recognition. However, this palm vein recognition model requires a large number of samples, a long training cycle, a strong computing platform, and, in most scenarios, an operating system (for example, Windows operating system) for running artificial neural networks. As a result, the hardware and system development costs are high, which is not conducive to promotion.
[0035] The present invention proposes a solution, the basic concept of which is to pool the grayscale image of the palm vein while retaining the grayscale information, reduce the total data dimension, and then based on the limited pooled data blocks, partially extract the data and send it to the identification generation model to establish a relationship mapping between the palm vein image and the palm identification code. In other words, it is to build a model that performs multiple classifications through the palm vein image and gives a classification number (identification code). Since the classification model structure and parameters are clear, it has less dependence on the software and hardware platform, and compared with complex algorithms such as artificial neural networks, the classification model structure is more compact, consumes less resources, and is easy to miniaturize and promote.
[0036] To achieve the above purpose, the embodiment of the present invention adjusts the grayscale interval of the obtained palm vein image, and extracts the edge features based on the grayscale to form an edge feature map. The acquisition of the palm vein image may be affected by various factors, resulting in changes in the overall or local brightness of the image. Moreover, since the embodiment of the present invention is based on the grayscale image rather than the black and white image for processing, the brightness of the image will affect the subsequent data processing results, or increase the difficulty of data processing. Therefore, the embodiment of the present invention first adjusts the grayscale image so that the pixel value distribution conforms to the preset distribution law. Specifically, all pixels are counted according to the predetermined interval, and the number obtained by counting is arranged to obtain a reference queue. For example, 0-255, a pixel value interval is set every 16 intervals, so that 16 intervals are obtained, and then each interval determines the number of pixel values (usually referring to the number of pixel values in each interval of the template), and these numbers are arranged according to the size of the interval to obtain a reference queue. That is, the element of each queue of the reference queue corresponds to an interval.
[0037] For the processing of the palm vein map, the pixel values of the palm vein map are first arranged according to size, and then the pixel values of the reference queue number are sequentially extracted from the arranged pixel value queue, and the pixel values in the palm vein map are adjusted with reference to the interval of the reference queue. In one method, the first formula is applied:
[0038] In the formula, is the pixel value after grayscale interval adjustment, is the pixel value before grayscale interval adjustment, is the minimum pixel value in the pixel set, is the maximum pixel value in the pixel set, is the upper limit of the pixel value interval, is the lower limit of the pixel value range.
[0039] In this way, the distribution interval of the pixel values of the palm vein image is unified with the target interval. On this basis, the embodiment of the present invention further extracts the edge of the palm vein grayscale image through a difference algorithm to obtain an edge feature map.
[0040] One way to extract edge feature maps using the difference method is to search for the minimum pixel value within the radius of 1 pixel for each pixel in the edge feature map, and replace the current value with the minimum value to obtain an edge convex map. For example, Figure 2 In the figure shown, the pixel 201 at the center position has a radius of 1 pixel and a total of 8 adjacent pixels, among which the pixel value above it is the smallest (132). At this time, the pixel value of this pixel is replaced by the pixel value of the pixel above it. After all pixels are processed as above, an edge convex image is obtained. This convex image and the palm vein grayscale image are subjected to edge extraction using the second formula to obtain an edge feature map. The second formula is:
[0041] In the formula, is the pixel value of the edge feature map, is the pixel value of the edge convex image, is the pixel value of the grayscale image.
[0042] In step 103, each edge feature map is pooled according to the first pooling degree parameter, and an identification code is assigned to the pooled data block obtained by pooling according to the palm corresponding to the palm vein map, wherein the pooled data blocks corresponding to the same palm correspond to the same identification code.
[0043] In some implementations, step 103 includes: For each edge feature map, perform the following steps: Obtaining a first pooling degree parameter, wherein the first pooling degree parameter includes: a pooling dimension and an offset; According to the first position, take out the data block of the pooling dimension from the edge feature map as the data block to be processed; Taking the maximum value or average value of multiple data in the data block to be processed as pooling data; Taking the first position as an index, adding the pooled data to a pooled data block; If the traversal of the edge feature map is not completed, the first position is offset according to the offset, and the process jumps to the step of taking out the data block of the pooling dimension from the edge feature map according to the first position as the data block to be processed.
[0044] Exemplarily, in order to refine the key features in the edge feature map, reduce the number of parameters input into the model, and reduce the complexity of the model, the embodiment of the present invention performs pooling on the edge feature map after obtaining the edge feature map.
[0045] like Figure 3 As shown, one pooling method is to extract data blocks (data blocks composed of pixel values) from the edge feature map according to a preset dimension, and put the maximum value in the extracted data block or the average value of the entire data block into the pooled data block according to the position in the edge feature map. In this way, each edge feature map extracts a pooled data block. On this basis, an identification code is assigned to each pooled data block. In fact, since the pooled data block comes from the palm vein map, and one palm usually corresponds to multiple palm vein maps (multiple collections of one palm), there will be a situation where multiple pooled data blocks correspond to one identification code. It should be noted that the correspondence between the palm and the identification code is unique, that is, different palms correspond to different identification codes.
[0046] In terms of setting the identification code, there are usually large differences in the identification codes corresponding to different palms; for example, in one application scenario, the identification code is a 9-digit identification code, in which the last three digits are the same, for example, the last three digits are all 0, and different palms are distinguished by the first 6 digits. The advantage of this is that when the identification code generation model is constructed later, it is not necessary to generate a strict identification code, and a certain amount of error tolerance is left. For example, when the model is built, when the user uses the model, the identification code recognized by the model through the palm vein map is xxxxxx200. Since the distinction interval between two adjacent identification codes is 1000, it can be seen that the error this time is 200, which is within an acceptable range. Then take the first 6 digits of the identification code as the identification code indication output by the model for identity confirmation; if it is xxxxxx600, it obviously means that the error is too large, the recognition fails this time, and the palm vein map needs to be obtained again.
[0047] That is to say, in terms of identification code setting, the embodiment of the present invention sets different identification codes for different palms and reserves an error identification area, so that the identification code indication output by the usage model can be compared with the preset identification code to determine the recognition error.
[0048] In step 104, multiple sub-data blocks are extracted from the pooled data block and input into the identification generation model. According to the deviation between the identification generated by the model and the identification of the pooled data block, the parameters of the identification generation model are adjusted until the deviation is less than a threshold, thereby obtaining a palm vein recognition model.
[0049] In some implementations, the identification generation model is:
[0050] In the formula, To identify the output, is the total number of rows of intermediate nodes, For the The weight coefficients of the output nodes, For the Line The first column node The weight coefficients, For the Line The output of the middle node of the column, For the input variables, is the total number of columns of intermediate nodes, is a natural constant, is the bias coefficient.
[0051] In some implementations, extracting a plurality of sub-data blocks from the pooled data block and inputting them into the identifier generation model includes: For each pooled data block, perform the following steps: Get the segmentation dimension and number of sub-blocks; Splitting the pooled data block into a plurality of sub-data blocks according to the segmentation dimension; Randomly selecting the number of sub-data blocks from the multiple sub-data blocks as multiple sub-data blocks to be processed; The multiple sub-data blocks to be processed are input into the identification generation model.
[0052] In some implementations, adjusting the parameters of the identifier generation model according to the deviation between the identifier generated by the model and the identifier of the pooled data block until the deviation is less than a threshold comprises: Acquire a plurality of coefficient arrays, wherein each coefficient array includes data corresponding to a plurality of coefficients in the identification generation model; Substituting the plurality of coefficient arrays into the identification generation model respectively to obtain a plurality of identification code indications, wherein each identification code indication corresponds to a coefficient array; Determine a plurality of identification deviations according to a target identification code and the plurality of identification code indications, wherein each identification deviation corresponds to a coefficient array, and the target identification code is an identification code of a pooled data block from which the plurality of sub-data blocks are derived; If there is a coefficient array whose identification deviation is greater than the deviation threshold, three optimal coefficient arrays are selected from the multiple coefficient arrays according to the multiple identification deviations, and the coefficient array greater than the deviation threshold is adjusted according to a third formula and the three optimal coefficient arrays, wherein the third formula is:
[0053] In the formula, For the The adjusted coefficient array data, For the The adjusted coefficient array data, , , as well as They are the first distance, the second distance, the third distance and the fourth distance respectively. , , as well as are the first coefficient, the second coefficient, the third coefficient and the fourth coefficient respectively, , , as well as are the optimal coefficient arrays. data, the suboptimal coefficient array data, the third best coefficient array data and the optimal coefficient array of the current coefficient array for each iteration individual data; If the traversal of all the pooled data blocks is not completed, multiple sub-data blocks are extracted from the pooled data blocks and input into the identification generation model, and the step of determining multiple identification deviations according to the target identification code and the multiple identification code indications is jumped to; If the traversal of all the pooled data blocks is completed and there is an unadjusted coefficient array in the current traversal process, the identification generation model into which the unadjusted coefficient array is substituted is used as the palm vein recognition model; Otherwise, a new round of traversal of the pooled data blocks is enabled, multiple sub-data blocks are extracted from the traversed pooled data blocks and input into the identification generation model, and the process jumps to the step of determining multiple identification deviations according to the target identification code and the multiple identification code indications.
[0054] Exemplarily, the embodiment of the present invention is to divide the pooled data block of the palm vein map into multiple sub-blocks, select some sub-blocks from the multiple sub-blocks and input them into the identification generation model, and adjust the parameters of the model according to the deviation of the model output and the identification code corresponding to the pooled data block. In the embodiment of the present invention, the identification generation model is:
[0055] In the formula, To identify the output, is the total number of rows of intermediate nodes, For the The weight coefficients of the output nodes, For the Line The first column node The weight coefficients, For the Line The output of the middle node of the column, For the input variables, is the total number of columns of intermediate nodes, is a natural constant, is the bias coefficient.
[0056] Based on this model, the embodiment of the present invention randomly selects several of the multiple sub-blocks obtained by segmenting the pooled data block and sends them to the model. After that, multiple coefficient arrays are respectively substituted into the model, that is, models with different coefficient arrays are obtained, and the identification code indications output by these models are compared with the identification code of the pooled data block to obtain the identification deviation, select the three best coefficient arrays, and adjust the other coefficient arrays using the third formula. The third formula is:
[0057] In the formula, For the The adjusted coefficient array data, For the The adjusted coefficient array data, , , as well as They are the first distance, the second distance, the third distance and the fourth distance respectively. , , as well as are the first coefficient, the second coefficient, the third coefficient and the fourth coefficient respectively, , , as well as are the optimal coefficient arrays. data, the suboptimal coefficient array data, the third best coefficient array data and the optimal coefficient array of the current coefficient array for each iteration data.
[0058] After the above adjustment, multiple sub-data blocks are extracted from the next pooled data block and input into the model to perform the above iteration. If, after traversing the pooled data blocks, there is a coefficient array that always makes the deviation less than the deviation threshold, then the iteration is successful. At this time, this array is used as the coefficient of the model and substituted into the model to obtain the palm vein recognition model.
[0059] In some scenarios, satisfactory iterative results may not be obtained after multiple rounds of iterations. In this case, it is necessary to adjust the number of sub-data blocks input to the model and the structure of the identification generation model. In other scenarios, the number of parameters input to the model can be adjusted by reducing the pooling depth to increase the dimension of the pooled data block and thus realize model construction.
[0060] In the application of the model, when a new palm vein map needs to be entered, the pooled data of the palm vein map is sent to the palm vein recognition model. At this time, the model will give an identification code (in the above example, the last three digits of the identification code will be set to zero), and the system will register this identification code. After the palm is entered, when identification is required, the pooled data of the palm vein map obtained is sent to the palm vein recognition model, and the model will give an identification code indication. This indication will be matched with the system's identification code (in the above example, the first 6 digits are matched), and when the match is successful, the accuracy is verified by the last three digits. When the verification is successful, the identity is confirmed, otherwise, it will request to obtain the palm vein map again. In other words, by obtaining the identification code indication, the double confirmation of identity authentication and verification accuracy is completed, and the security and reliability are relatively high.
[0061] The palm vein recognition model construction method of the present invention is implemented in such a way that first a plurality of palm vein images are obtained; then the plurality of palm vein images are grayscale adjusted and edge features are extracted to obtain a plurality of edge feature images, wherein each edge feature image corresponds to a palm vein image; then each edge feature image is pooled according to a first pooling degree parameter, and an identification code is assigned to the pooled data block obtained by pooling according to the palm corresponding to the palm vein image, wherein the pooled data blocks corresponding to the same palm correspond to the same identification code; finally, a plurality of sub-data blocks are extracted from the pooled data block and input into the identification generation model, and the parameters of the identification generation model are adjusted according to the deviation between the identification generated by the model and the identification of the pooled data block until the deviation is less than a threshold value, thereby obtaining a palm vein recognition model. The implementation method of the present invention extracts edge features after grayscale processing of the palm vein image, and contains richer feature content than the edge features extracted by binarization, which is convenient for improving the recognition rate; the implementation method of the present invention extracts part of the data in the pooled data block after pooling and sends it to the model, and performs identification code fitting on the model, so that when the model is applied, only a small part of the palm vein image needs to be obtained to complete the recognition, thereby improving the recognition efficiency and accuracy. The implementation mode of the present invention does not require an intelligent algorithm. Instead, the model parameters are determined by generating multiple sub-data blocks through a palm vein map. Compared with traditional intelligent algorithms, it requires fewer samples, has low dependence on software platforms and hardware platforms, consumes less resources in both early modeling and later use, and is easy to use and promote.
[0062] It should be understood that the size of the serial numbers of the steps in the above implementation does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the implementation method of the present invention.
[0063] The following is an implementation of the device of the present invention. For details not described in detail, reference may be made to the corresponding method implementation described above.
[0064] Figure 4 is a functional block diagram of a palm vein recognition model building device provided by an embodiment of the present invention, referring to Figure 4 The palm vein recognition model building device includes: an image acquisition module 401, an edge extraction module 402, a pooling module 403 and a model building module 404, wherein: The image acquisition module 401 is used to acquire multiple palm vein images; The edge extraction module 402 is used to adjust the grayscale interval and extract edge features of the plurality of palm vein images to obtain a plurality of edge feature images, wherein each edge feature image corresponds to a palm vein image; A pooling module 403 is used to pool each edge feature map according to a first pooling degree parameter, and assign an identification code to a pooled data block obtained by pooling according to the palm corresponding to the palm vein map, wherein the pooled data blocks corresponding to the same palm correspond to the same identification code; as well as, The model building module 404 is used to extract multiple sub-data blocks from the pooled data block and input them into the identification generation model, and adjust the parameters of the identification generation model according to the deviation between the identification generated by the model and the identification of the pooled data block until the deviation is less than a threshold, so as to obtain a palm vein recognition model.
[0065] Figure 5 is a functional block diagram of an electronic device provided by an embodiment of the present invention. Figure 5 As shown, the electronic device 5 of this embodiment includes: a processor 500 and a memory 501, wherein the memory 501 stores a computer program 502 that can be run on the processor 500. When the processor 500 executes the computer program 502, the steps in the above-mentioned palm vein recognition model construction method and the embodiment are implemented, for example Figure 1 Steps 101 to 104 are shown.
[0066] Exemplarily, the computer program 502 may be divided into one or more modules / units, and the one or more modules / units are stored in the memory 501 and executed by the processor 500 to implement the present invention.
[0067] The electronic device 5 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The electronic device 5 may include, but is not limited to, a processor 500 and a memory 501. Those skilled in the art will appreciate that Figure 5 It is only an example of the electronic device 5 and does not constitute a limitation of the electronic device 5. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device 5 may also include input and output devices, network access devices, buses, etc.
[0068] The processor 500 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.
[0069] The memory 501 may be an internal storage unit of the electronic device 5, such as a hard disk or memory of the electronic device 5. The memory 501 may also be an external storage device of the electronic device 5, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 5. Further, the memory 501 may also include both an internal storage unit of the electronic device 5 and an external storage device. The memory 501 is used to store the computer program 502 and other programs and data required by the electronic device 5. The memory 501 may also be used to temporarily store data that has been output or is to be output.
[0070] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the implementation method can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method implementation method, which will not be repeated here.
[0071] In the above-mentioned embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0072] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0073] In the embodiments provided by the present invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0074] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.
[0075] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0076] If the integrated module / unit is implemented in the form of 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 present invention implements all or part of the processes in the above-mentioned implementation method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various methods and device implementation methods can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium.
[0077] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the protection scope of the present invention.
Claims
1. A palm vein recognition model construction method, characterized in that: include: Obtain multiple palm vein maps; Performing grayscale interval adjustment and edge feature extraction on the multiple palm vein images to obtain multiple edge feature images, wherein each edge feature image corresponds to a palm vein image; Pooling each edge feature map according to the first pooling degree parameter, and assigning an identification code to a pooled data block obtained by pooling according to the palm corresponding to the palm vein map, wherein the pooled data blocks corresponding to the same palm correspond to the same identification code; A plurality of sub-data blocks are extracted from the pooled data block and input into the identification generation model. According to the deviation between the identification generated by the model and the identification of the pooled data block, the parameters of the identification generation model are adjusted until the deviation is less than a threshold, thereby obtaining a palm vein recognition model.
2. The palm vein recognition model construction method according to claim 1, characterized in that: The grayscale interval adjustment and edge feature extraction of the plurality of palm vein images to obtain a plurality of edge feature images includes: For each of the multiple palm vein maps, the following steps are performed respectively: Acquire a reference queue and a plurality of pixel value intervals, wherein each queue element of the reference queue corresponds to a pixel value interval; Arrange multiple pixel values in the palm vein image according to the size of the values to obtain a pixel value queue; Extracting pixel values from the pixel value queue according to the reference queue to construct a plurality of pixel sets, wherein each pixel set corresponds to a queue element of the reference queue; According to the multiple pixel sets and the multiple pixel value intervals, adjusting the grayscale intervals of the pixels in the palm vein image to obtain a grayscale image; The edge of the grayscale image is extracted using a difference method to obtain an edge feature map.
3. The palm vein recognition model construction method according to claim 2, characterized in that: The step of adjusting the grayscale intervals of pixels in the palm vein image according to the multiple pixel sets and the multiple pixel value intervals to obtain a grayscale image includes: According to the first formula, the plurality of pixel sets and the plurality of pixel value intervals, the grayscale intervals of the pixels in the palm vein image are adjusted to obtain a grayscale image, wherein the first formula is: In the formula, is the pixel value after grayscale interval adjustment, is the pixel value before grayscale interval adjustment, is the minimum pixel value in the pixel set, is the maximum pixel value in the pixel set, is the upper limit of the pixel value interval, is the lower limit of the pixel value interval; The step of extracting edges from the grayscale image using a differential method to obtain an edge feature map includes: For each pixel in the edge feature map, a minimum pixel value within a radius of 1 pixel is used to replace the pixel value to obtain an edge convexification map; The edge of the grayscale image is extracted according to a second formula, the grayscale image and the edge convexification image, wherein the second formula is: In the formula, is the pixel value of the edge feature map, is the pixel value of the edge convex image, is the pixel value of the grayscale image.
4. The palm vein recognition model construction method according to claim 1, characterized in that: The step of pooling each edge feature map according to the first pooling degree parameter includes: For each edge feature map, perform the following steps: Obtaining a first pooling degree parameter, wherein the first pooling degree parameter includes: a pooling dimension and an offset; According to the first position, take out the data block of the pooling dimension from the edge feature map as the data block to be processed; Taking the maximum value or average value of multiple data in the data block to be processed as pooling data; Taking the first position as an index, adding the pooled data to a pooled data block; If the traversal of the edge feature map is not completed, the first position is offset according to the offset, and the process jumps to the step of taking out the data block of the pooling dimension from the edge feature map according to the first position as the data block to be processed.
5. The palm vein recognition model construction method according to any one of claims 1 to 4, characterized in that: The identification generation model is: In the formula, To identify the output, is the total number of rows of intermediate nodes, For the The weight coefficients of the output nodes, For the Line The first column node The weight coefficients, For the Line The output of the middle node of the column, For the input variables, is the total number of columns of intermediate nodes, is a natural constant, is the bias coefficient.
6. The palm vein recognition model construction method according to claim 5, characterized in that: The step of extracting a plurality of sub-data blocks from the pooled data block and inputting the sub-data blocks into the identifier generation model comprises: For each pooled data block, perform the following steps: Get the segmentation dimension and number of sub-blocks; Splitting the pooled data block into a plurality of sub-data blocks according to the segmentation dimension; Randomly selecting the number of sub-data blocks from the multiple sub-data blocks as multiple sub-data blocks to be processed; The multiple sub-data blocks to be processed are input into the identification generation model.
7. The palm vein recognition model construction method according to claim 6, characterized in that: The step of adjusting the parameters of the identifier generation model according to the deviation between the identifier generated by the model and the identifier of the pooled data block until the deviation is less than a threshold comprises: Acquire a plurality of coefficient arrays, wherein each coefficient array includes data corresponding to a plurality of coefficients in the identification generation model; Substituting the plurality of coefficient arrays into the identification generation model respectively to obtain a plurality of identification code indications, wherein each identification code indication corresponds to a coefficient array; Determine a plurality of identification deviations according to a target identification code and the plurality of identification code indications, wherein each identification deviation corresponds to a coefficient array, and the target identification code is an identification code of a pooled data block from which the plurality of sub-data blocks are derived; If there is a coefficient array whose identification deviation is greater than the deviation threshold, three optimal coefficient arrays are selected from the multiple coefficient arrays according to the multiple identification deviations, and the coefficient array greater than the deviation threshold is adjusted according to a third formula and the three optimal coefficient arrays, wherein the third formula is: In the formula, For the The adjusted coefficient array data, For the The adjusted coefficient array data, , , as well as They are the first distance, the second distance, the third distance and the fourth distance respectively. , , as well as are the first coefficient, the second coefficient, the third coefficient and the fourth coefficient respectively, , , as well as are the optimal coefficient arrays. data, the suboptimal coefficient array data, the third best coefficient array data and the optimal coefficient array of the current coefficient array for each iteration individual data; If the traversal of all the pooled data blocks is not completed, multiple sub-data blocks are extracted from the pooled data blocks and input into the identification generation model, and the step of determining multiple identification deviations according to the target identification code and the multiple identification code indications is jumped to; If the traversal of all the pooled data blocks is completed and there is an unadjusted coefficient array in the current traversal process, the identification generation model into which the unadjusted coefficient array is substituted is used as the palm vein recognition model; Otherwise, a new round of traversal of the pooled data blocks is enabled, multiple sub-data blocks are extracted from the traversed pooled data blocks and input into the identification generation model, and the process jumps to the step of determining multiple identification deviations according to the target identification code and the multiple identification code indications.
8. A palm vein recognition model construction device, characterized in that: Used to implement the palm vein recognition model construction method according to any one of claims 1 to 7, the palm vein recognition model construction device comprises: An image acquisition module, used to acquire multiple palm vein images; An edge extraction module, used for performing grayscale interval adjustment and edge feature extraction on the plurality of palm vein images to obtain a plurality of edge feature images, wherein each edge feature image corresponds to a palm vein image; A pooling module, used for pooling each edge feature map according to a first pooling degree parameter, and assigning an identification code to a pooled data block obtained by pooling according to a palm corresponding to the palm vein map, wherein pooled data blocks corresponding to the same palm correspond to the same identification code; as well as, The model building module is used to extract multiple sub-data blocks from the pooled data block and input them into the identification generation model. According to the deviation between the identification generated by the model and the identification of the pooled data block, the parameters of the identification generation model are adjusted until the deviation is less than a threshold, thereby obtaining a palm vein recognition model.
9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method as claimed in any one of claims 1 to 7 are implemented.
Citation Information
Patent Citations
Non-contact biological recognition guiding device and method and biological feature recognition system
CN111881719A
Palm vein image anti-counterfeiting method, device and equipment
CN112861743A
Gate control system and method based on palm vein technology
CN116030503A
Palm vein recognition method, device and equipment and storage medium
CN116052230A
Photovoltaic power generation power prediction model construction method and device
CN117454180A
