Palmprint image generation method, device and equipment based on B-spline curve and CNN

Abstract

The invention provides a palm print image generation method, device and equipment based on B-spline curve and CNN, which relate to the technical field of biometric feature recognition. Firstly, a suitable area is selected to simulate a region of interest (ROI), and at the same time, the ROI area is divided. After the division is completed, corresponding parameter points are selected in the corresponding area; then, a palm print line image is fitted using a B-spline curve, wherein each palm print line image corresponds to a different user; finally, a CNN is used to fuse the features of a real palm print image with the palm print line image to form a pseudo palm print database that is closer to the real palm print, thereby solving the problem of small data volume of the palm print database in the current palm print recognition research.

Description

Technical Field

[0001] The present application relates to the field of biometric feature recognition, and in particular to a palmprint image generation method, device and equipment based on B-spline curve and CNN. Background Art

[0002] Biometric recognition technology is a technology that uses measurable physiological or behavioral characteristics of the human body to identify and verify identity. This technology has been widely used in finance, medical care, education, transportation, security and other fields. Among these technologies, fingerprint recognition, face recognition, palm print recognition, iris recognition, gait recognition and voice print recognition are common. Each recognition technology has its advantages and disadvantages. Fingerprint recognition is limited by its small area, insufficient information richness and contact collection method that is prone to deformation and hygiene problems. Face recognition is plagued by problems such as large occlusion interference, easy to be stolen, and reliability needs to be improved. Palm print images have highly recognizable texture features, including main lines, wrinkles, ridge ends and bifurcation points, and have high recognition accuracy. In addition, palm print images are easy to collect, even with low-resolution mobile phones, and are stable and reliable. Therefore, compared with fingerprint recognition and face recognition, palm print feature recognition has the advantages of rich features, easy user acceptance, safety and stability, and is the future development trend of biometric recognition.

[0003] Recognition algorithms based on deep learning can learn more complex filter structures from data by themselves. The extracted palmprint depth features are richer, more robust, and have stronger generalization capabilities. However, the training of neural networks usually requires a large amount of data. The current public palmprint datasets have the problems of small numbers of individuals and insufficient palmprint images of a single individual. Summary of the invention

[0004] In order to solve the problem of small data volume in palmprint database in current palmprint recognition research, the present application provides a palmprint image generation method, device and equipment based on B-spline curve and CNN.

[0005] In the first aspect, the present application provides a palmprint image generation method based on B-spline curve and CNN, which adopts the following technical solution:

[0006] A palmprint image generation method based on B-spline curve and CNN, comprising:

[0007] According to the prior knowledge, the structural features of the palm print are obtained, a square area of ​​a preset size is selected as an ROI area of ​​interest for simulating the palm print according to the structural features, and the ROI area is divided into regions;

[0008] In different areas after division, parameterized B-spline curves are selected to simulate relevant parameter variables in the palm lines in different areas of ROI;

[0009] Using a B-spline curve to fit the palm lines according to the ROI area and the relevant parameter variables in the corresponding area to obtain a generated palm line image;

[0010] Extracting a first image feature from the generated palm line image and a second image feature from the real palm line image using a convolutional neural network (CNN) model;

[0011] The first image feature and the second image feature are feature-fused to generate a pseudo palmprint image.

[0012] Optionally, the preset size is 7 cm×7 cm, and dividing the ROI area into regions includes:

[0013] Divide the ROI area into four square areas, wherein each of the divided square areas is uniform and has the same size, and the four square areas are respectively an upper left area, a lower left area, an upper right area, and a lower right area;

[0014] The upper left area among the four square areas is selected, and the upper left area is also divided into four square areas in the same manner.

[0015] Optionally, the relevant parameter variables include endpoints and control points, the endpoints include the starting point and the end point of the main line, and the selecting and using parameterized B-spline curves in different areas of the ROI after division to simulate the relevant parameter variables in the palm line includes:

[0016] Selecting the starting point of the main pattern from the four square areas divided by the upper left area, and using the starting point as a reference to parameterize the palm crease corresponding to the main pattern using a B-spline curve;

[0017] Selecting parameter points from the lower left area, the upper right area and the lower right area as the end points of the main lines based on the palm crease;

[0018] A connecting line is established between the starting point and the end point, and a number of control points are selected from the area formed between the connecting line and the palm crease corresponding to the main pattern.

[0019] Optionally, the using a B-spline curve to fit the palm line according to the ROI area and the relevant parameter variables in the corresponding area to obtain a generated palm line image includes:

[0020] The points on the B-spline curve are calculated according to the endpoints and the control points, and the calculation formula is:

[0021] ;

[0022] Among them, Ci i = (0,1,...,n) is the control point, N i,k (t)(i = 1,2,...,n) is the basis function of the B-spline curve, and the recursive formula is:

[0023] When k=0, , k=0;

[0024] When k ≥ 1, , i = 0,1,...,n;

[0025] Where k is the degree of the B-spline basis function, and U = is the endpoint vector.

[0026] Optionally, the step of fusing the first image feature with the second image feature to generate a pseudo palmprint image includes:

[0027] Initialize a style image and a content image based on the first image feature and the second image feature;

[0028] Define a loss function, wherein the loss function includes content loss, style loss, and total variation loss;

[0029] Selecting a feature extraction part in the CNN model, using five convolutions to output style features based on the style image and the style loss, and using one convolution layer to output content features based on the content image and the content loss;

[0030] A pseudo palmprint image is generated based on the style feature, the content feature and the total variation loss.

[0031] Optionally, the content loss is used to measure the difference between the generated image and the content image, and the calculation formula is:

[0032] ;

[0033] in and are the original content image and the generated image, It is The activation value of the ith filter at position j in the layer, It is The original value of the i-th filter at position j in the layer;

[0034] The style loss is used to measure the difference between the generated image and the style image, and the calculation formula is:

[0035] ;

[0036] in, and are the original style image and the generated image, is the weight factor of each layer’s contribution to the total loss;

[0037] The total variation loss is used to reduce the noise in the generated image, and the calculation formula is:

[0038] ;

[0039] Among them, α and β are the weight factors for content and style reconstruction, respectively.

[0040] Optional, expected The calculation formula is:

[0041] ;

[0042] in, Represents the number of feature maps, and the size of each feature map is ,in is the height of the feature map multiplied by its width, It is The inner product between the vectorized feature maps i and j in the layer, .

[0043] In a second aspect, the present application provides a palmprint image generation device based on a B-spline curve and a CNN, the palmprint image generation device based on a B-spline curve and a CNN comprising:

[0044] A division module is used to obtain the structural features of the palm print according to prior knowledge, select a square area of ​​a preset size as an interested ROI area of ​​the simulated palm print according to the structural features, and divide the ROI area into regions;

[0045] A selection module is used to select and use parameterized B-spline curves to simulate relevant parameter variables in the palm lines in different areas of the ROI after division;

[0046] A fitting module, used for fitting the palm lines using a B-spline curve according to the ROI area and the relevant parameter variables in the corresponding area to obtain a generated palm line image;

[0047] An extraction module, used for extracting a first image feature from the generated palm line image and a second image feature from the real palm line image using a convolutional neural network (CNN) model;

[0048] The fusion module is used to fuse the first image feature and the second image feature to generate a pseudo palmprint image.

[0049] Optionally, the preset size is 7 cm×7 cm, and the division module is used to divide the ROI area into four square areas, wherein each of the divided square areas is uniform and has the same size, and the four square areas are respectively an upper left area, a lower left area, an upper right area, and a lower right area;

[0050] The upper left area among the four square areas is selected, and the upper left area is also divided into four square areas in the same manner.

[0051] In a third aspect, the present application provides a palmprint image generation device based on B-spline curves and CNN, the device comprising: a memory, a processor, and the processor executes the method described above when running the palmprint image generation program based on B-spline curves and CNN stored in the memory.

[0052] In summary, the present application includes the following beneficial technical effects:

[0053] The present application obtains the structural features of palmprints according to prior knowledge, selects a square area of ​​a preset size as an interested ROI area for simulating palmprints according to the structural features, and divides the ROI area into regions; selects and uses parameterized B-spline curves in different regions of the ROI after division to simulate relevant parameter variables in the palmprint line; uses the B-spline curve to fit the palmprint line according to the ROI area and the relevant parameter variables in the corresponding area to obtain a generated palmprint line image; uses a convolutional neural network (CNN) model to extract a first image feature from the generated palmprint line image and a second image feature from the real palmprint image; performs feature fusion on the first image feature and the second image feature to generate a pseudo palmprint image. The pseudo palmprint image is generated in the above manner for use in the subsequent construction of a palmprint database, thereby solving the problem of small data volume in the palmprint database in current palmprint recognition research. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a schematic diagram of the structure of a palmprint image generation device based on B-spline curve and CNN in the hardware operating environment involved in the embodiment of the present application;

[0055] Figure 2 It is a flowchart of the first embodiment of the palmprint image generation method based on B-spline curve and CNN of the present application;

[0056] Figure 3 It is a schematic diagram of the overall framework flow of the palmprint image generation method based on B-spline curve and CNN in the palmprint image generation method based on B-spline curve and CNN of the present application;

[0057] Figure 4This is a schematic diagram of ROI area division in the palmprint image generation method based on B-spline curve and CNN in this application;

[0058] Figure 5 It is a schematic diagram of the main pattern of the real palm print in the palm print image generation method based on B-spline curve and CNN in this application;

[0059] Figure 6 It is a schematic diagram of the main pattern of the simulated palm print in the palm print image generation method based on B-spline curve and CNN of the present application;

[0060] Figure 7 It is a schematic diagram of the distribution of control points in the palmprint image generation method based on B-spline curve and CNN in this application;

[0061] Figure 8 It is a schematic diagram of palmprint synthesis based on CNN in the palmprint image generation method based on B-spline curve and CNN of the present application;

[0062] Fig. 9 It is a structural block diagram of the first embodiment of the palmprint image generation device based on B-spline curve and CNN of the present application. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below through the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0064] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a palmprint image generation device based on B-spline curves and CNN in the hardware operating environment involved in the embodiment of the present application.

[0065] like Figure 1As shown, the palm print image generation device based on B-spline curve and CNN may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (Wireless-Fidelity, Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM), or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk storage. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0066] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation on the palmprint image generating device based on B-spline curve and CNN, and may include more or less components than shown in the figure, or combine some components, or arrange the components differently.

[0067] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a palm print image generation program based on B-spline curves and CNN.

[0068] exist Figure 1 In the palmprint image generation device based on B-spline curve and CNN shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the present application can be set in the palmprint image generation device based on B-spline curve and CNN, and the palmprint image generation device based on B-spline curve and CNN calls the palmprint image generation program based on B-spline curve and CNN stored in the memory 1005 through the processor 1001, and executes the palmprint image generation method based on B-spline curve and CNN provided in the embodiment of the present application.

[0069] The present application embodiment provides a palmprint image generation method based on B-spline curve and CNN, referring to Figure 2 , Figure 2This is a flow chart of the first embodiment of the palmprint image generation method based on B-spline curve and CNN of the present application.

[0070] The overall process of this program can refer to Figure 3 As shown, first, the basic structural features of the palm print are obtained according to the prior knowledge, and the region of interest (ROI) of the palm print is simulated by selecting a suitable square area through the structural features. At the same time, the ROI area is divided into regions, and the parameterized B-spline curve is determined in different regions of the ROI to simulate the relevant parameter variables (endpoints and control points) in the palm print line; then, the palm print line is simulated by randomly selecting parameter variables in the specified area, thereby realizing the generation of multiple users (IDs), and there are also inter-class differences in the same ID. This change is used to simulate the differences caused by some external interference in the process of collecting palm print images; finally, in order to make the generated palm print line image similar to the real palm print image, the convolutional neural network (CNN) model is used to extract features from the real palm print image and the generated palm print line image. After the extraction is completed, the features of the two types of images are fused to generate a palm print image that is extremely similar to the real palm print. This embodiment can well solve the problem of small data volume of palm print database in the current palm print recognition research, and has strong practical value.

[0071] Step S10: obtaining the structural features of the palmprint according to prior knowledge, selecting a square area of ​​a preset size as an ROI area of ​​interest for simulating the palmprint according to the structural features, and dividing the ROI area.

[0072] In the specific implementation, in this embodiment, it is necessary to first obtain the structural characteristics of the palm print according to prior knowledge, and then divide the ROI area. It should be noted that the size of an adult's palm corresponds to gender and height, but for the use of a model for recognition, the key lies in the palm print line structure of the ROI area on the palm. The size of the palm does not affect the recognition work, so a square area of ​​7 cm × 7 cm is uniformly selected as the ROI area of ​​the synthetic palm print image.

[0073] The division process is specifically to divide the ROI area into four square areas, wherein each of the divided square areas is uniform and of the same size, and the four square areas are respectively an upper left area, a lower left area, an upper right area and a lower right area; the upper left area is selected from the four square areas, and the upper left area is also divided into four square areas in the same way.

[0074] The above division process can refer to Figure 4As shown, according to the structural characteristics of the real palmprint, the synthesized ROI area is further divided into four even and equal-sized small squares, namely the upper left, lower left, upper right, and lower right. At the same time, the same segmentation method is performed on the upper left area, namely left A, left D, right A, and right D.

[0075] Step S20: Selecting parameterized B-spline curves in different divided regions and different regions of the ROI to simulate relevant parameter variables in the palm lines.

[0076] It should be noted that the relevant parameter variables include endpoints and control points, and the endpoints include the starting point and end point of the main pattern. The above process specifically includes selecting the starting point of the main pattern from the four square areas divided by the upper left area, and using the B-spline curve to parameterize the palm crease corresponding to the main pattern based on the starting point; selecting parameter points from the lower left area, the upper right area and the lower right area as the end points of the main pattern based on the palm crease; establishing a connecting line between the starting point and the end point, and selecting a number of control points from the area formed between the connecting line and the palm crease corresponding to the main pattern. Figure 5 As shown in the figure, the real palm print image can be obtained. The palm print is roughly composed of several (usually 3 to 5) main lines and some fine wrinkles. The endpoints of the main lines are distributed in different areas. According to the above-mentioned divided areas, the main lines of the palm prints can be selected into the corresponding endpoints of the selected areas, and the palm creases are parameterized using B-spline curves. Taking the three main lines in the left palm as an example, the starting points of the three main lines are all located in the upper left area, but the starting point of Line1 is located in the right A area, the starting points of Line2 and Line3 are both located in the left A area and the two points are relatively close, and the end points are located in different areas (the end point of Line1 is located in the upper right area, the end point of Line2 is located in the lower right area, and the end point of Line3 is located in the lower left area). Among them, Line1~3 are palm creases, and the corresponding starting points and end points of Line1~3 are also the above-mentioned end points.

[0077] Furthermore, according to the observation of real palm lines, the distribution characteristics of the three main lines are as follows (all taking the left palm as an example): Line 1 is mainly convex downward, Line 3 is mainly convex upward, and Line 2 is relatively flat without obvious concave and convex characteristics. The simulated palm lines are as follows: Figure 6 As shown. According to the endpoints determined above, control points are further selected within a suitable range to make the main line layout of the B-spline curve fitting more similar to a real palm. For example, the control points in Line1 are mainly selected in the elliptical area below the endpoint line. Since Line2 has no obvious concave and convex characteristics, its control points are selected in the middle elliptical area of ​​the endpoint line. The control points in Line3 are mainly selected in the elliptical area above the endpoint line. The selection process of the above control points can refer to Figure 7 As shown, the number of control points selected is 2 to 4, the purpose is to achieve intra-class differences. In addition, in order to more realistically restore the impact of the external environment, random noise and fuzziness are added during the synthesis of palmprints to simulate the differences in the collected images caused by changes in the external environment under normal circumstances.

[0078] Step S30: using a B-spline curve to fit the palm lines according to the ROI area and the relevant parameter variables in the corresponding area to obtain a generated palm line image.

[0079] It should be noted that when using B-spline curve to fit palmprints, the target curve is divided into several parts due to the existence of endpoints. The purpose is to make each part have some influence but also a certain degree of independence, so that the change of one control point will not greatly affect the entire curve. This is very helpful to produce intra-class differences under the same ID (used to simulate image differences caused by the external environment).

[0080] The points on the B-spline curve are calculated according to the endpoints and the control points, and the calculation formula is:

[0081] ;

[0082] Among them, C i i = (0,1,...,n) is the control point, N i,k (t)(i = 1,2,...,n) is the basis function of the B-spline curve, and the recursive formula is:

[0083] When k=0, , k=0;

[0084] When k ≥ 1, , i = 0,1,...,n;

[0085] Where k is the degree of the B-spline basis function, and U = is the endpoint vector.

[0086] It should be further explained that there are many differences between male and female palms, such as palm color and texture. Male palms are usually reddish and partially darker. In addition, male grip strength is generally greater than female, so the palm lines are darker. Figure 6 The thickness of the target curve is used to synthesize female and male palm prints respectively, making the synthesis database more complete.

[0087] Step S40: extracting a first image feature from the generated palm line image and a second image feature from the real palm line image using a convolutional neural network (CNN) model.

[0088] Step S50: Fusing the first image feature and the second image feature to generate a pseudo palmprint image.

[0089] Specifically, a style image and a content image can be initialized based on the first image feature and the second image feature; a loss function is defined, wherein the loss function includes content loss, style loss, and total variation loss; a feature extraction part in the CNN model is selected, and style features are output based on the style image and the style loss using five convolutions, and content features are output based on the content image and the content loss using one convolution layer; and a pseudo palmprint image is generated based on the style features, the content features, and the total variation loss.

[0090] It should be noted that the loss function in this embodiment consists of three parts: content loss, style loss and total variation loss, where:

[0091] (1) Content loss is used to measure the difference between the generated image and the content image, making the content features close. The feature representation of a certain layer of the convolutional network is used to calculate the content loss, as shown in the following formula:

[0092] ;

[0093] in, and are the original content image and the generated image, It is The activation value of the ith filter at position j in the layer, It is The original value of the i-th filter at position j in the layer.

[0094] (2) Style loss is used to measure the difference between the generated image and the style image, making the style features close. The style loss is calculated by comparing the style features of a specific layer, as shown in the following formula:

[0095] ;

[0096] in and are the original style image and the generated image, is the weight factor of each layer's contribution to the total loss, and the expectation is As shown below:

[0097] ;

[0098] in, Represents the number of feature maps, and the size of each feature map is ,in is the height times the width of the feature map. The expectation depends on the spatial extent of the feature map. These feature correlations are represented by the Gram matrix ,in It is The inner product between the vectorized feature maps i and j in the layer is as follows:

[0099] ;

[0100] (3) Total variation loss is used to reduce the noise in the generated image and make the result smoother. The loss function is shown as follows:

[0101] ;

[0102] Among them, α and β are the weight factors for content and style reconstruction, respectively.

[0103] Furthermore, deep convolutional neural networks extract image features step by step with multiple layers, and the output of some layers can be selected as content features or style features. The feature extraction part of the CNN model is selected, in which 5 convolutional layers (conv1_1, conv2_1, conv3_1, conv4_1, conv5_1) output style features, and one convolutional layer (conv4_2) outputs content features. The loss function of feature fusion is calculated through forward propagation, and the model parameters are iterated through back propagation, that is, the synthetic image is continuously updated. The feature fusion process is as follows: Figure 8 shown.

[0104] In this embodiment, the structural features of the palm print are obtained according to prior knowledge, and a square area of ​​a preset size is selected as an interested ROI area for simulating the palm print according to the structural features, and the ROI area is divided into regions; in different regions after the division, a parameterized B-spline curve is selected to simulate relevant parameter variables in the palm print line; the palm print line is fitted according to the relevant parameter variables in the ROI area and the corresponding area using the B-spline curve to obtain a generated palm print line image; a convolutional neural network CNN model is used to extract a first image feature from the generated palm print line image and a second image feature from the real palm print image; the first image feature and the second image feature are feature-fused to generate a pseudo palm print image. The pseudo palm print image is generated in the above manner for use in the subsequent construction of a palm print database, which solves the problem of small data volume of the palm print database in the current palm print recognition research.

[0105] Reference Fig. 9 , Fig. 9 This is a structural block diagram of the first embodiment of the palmprint image generation device based on B-spline curve and CNN of the present application.

[0106] like Fig. 9As shown, the palmprint image generation device based on B-spline curve and CNN proposed in the embodiment of the present application includes:

[0107] The division module 10 is used to obtain the structural features of the palm print according to prior knowledge, select a square area of ​​a preset size as an interested ROI area of ​​the simulated palm print according to the structural features, and divide the ROI area into regions.

[0108] The selection module 20 is used to select and use parameterized B-spline curves to simulate relevant parameter variables in the palm lines in different areas of the ROI after the division.

[0109] The fitting module 30 is used to fit the palm lines using a B-spline curve according to the ROI area and the relevant parameter variables in the corresponding area to obtain a generated palm line image.

[0110] The extraction module 40 is used to extract the first image feature from the generated palm line image and the second image feature from the real palm line image using a convolutional neural network (CNN) model.

[0111] The fusion module 50 is used to fuse the first image feature and the second image feature to generate a pseudo palmprint image.

[0112] In this embodiment, the structural features of the palm print are obtained according to prior knowledge, and a square area of ​​a preset size is selected as an interested ROI area for simulating the palm print according to the structural features, and the ROI area is divided into regions; in different regions after the division, a parameterized B-spline curve is selected to simulate relevant parameter variables in the palm print line; the palm print line is fitted according to the relevant parameter variables in the ROI area and the corresponding area using the B-spline curve to obtain a generated palm print line image; a convolutional neural network CNN model is used to extract a first image feature from the generated palm print line image and a second image feature from the real palm print image; the first image feature and the second image feature are feature-fused to generate a pseudo palm print image. The pseudo palm print image is generated in the above manner for use in the subsequent construction of a palm print database, which solves the problem of small data volume of the palm print database in the current palm print recognition research.

[0113] In some embodiments, the preset size is 7 cm×7 cm, and the division module is used to divide the ROI area into four square areas, wherein each of the divided square areas is uniform and has the same size, and the four square areas are respectively an upper left area, a lower left area, an upper right area, and a lower right area;

[0114] The upper left area among the four square areas is selected, and the upper left area is also divided into four square areas in the same manner.

[0115] In some embodiments, the relevant parameter variables include endpoints and control points, the endpoints include the starting point and the end point of the main pattern, the selection module 20 is used to select the starting point of the main pattern from the four square areas divided by the upper left area, and use the B-spline curve to parameterize the palm crease corresponding to the main pattern based on the starting point;

[0116] Selecting parameter points from the lower left area, the upper right area and the lower right area as the end points of the main lines based on the palm crease;

[0117] A connecting line is established between the starting point and the end point, and a number of control points are selected from the area formed between the connecting line and the palm crease corresponding to the main pattern.

[0118] In some embodiments, the fitting module 30 is used to calculate the points on the B-spline curve according to the endpoints and the control points, and the calculation formula is:

[0119] ;

[0120] Among them, C i i = (0,1,...,n) is the control point, N i,k (t)(i = 1,2,...,n) is the basis function of the B-spline curve, and the recursive formula is:

[0121] When k=0, , k=0;

[0122] When k ≥ 1, , i = 0,1,...,n;

[0123] Where k is the degree of the B-spline basis function, and U = is the endpoint vector.

[0124] In some embodiments, the fusion module 50 is used to initialize the style image and the content image based on the first image feature and the second image feature;

[0125] Define a loss function, wherein the loss function includes content loss, style loss, and total variation loss;

[0126] Selecting a feature extraction part in the CNN model, using five convolutions to output style features based on the style image and the style loss, and using one convolution layer to output content features based on the content image and the content loss;

[0127] A pseudo palmprint image is generated based on the style feature, the content feature and the total variation loss.

[0128] In some embodiments, the content loss is used to measure the difference between the generated image and the content image, and the calculation formula is:

[0129] ;

[0130] in and are the original content image and the generated image, It is The activation value of the ith filter at position j in the layer, It is The original value of the i-th filter at position j in the layer;

[0131] The style loss is used to measure the difference between the generated image and the style image, and the calculation formula is:

[0132] in, and are the original style image and the generated image, is the weight factor of each layer’s contribution to the total loss;

[0133] The total variation loss is used to reduce the noise in the generated image, and the calculation formula is:

[0134] ;

[0135] Among them, α and β are the weight factors for content and style reconstruction, respectively.

[0136] In some embodiments, The expected calculation formula is:

[0137] ;

[0138] in, Represents the number of feature maps, and the size of each feature map is ,in is the height of the feature map multiplied by its width, It is The inner product between the vectorized feature maps i and j in the layer, .

[0139] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present application. In practical applications, technicians in this field can select part or all of it according to actual needs to achieve the purpose of the present embodiment, and no limitation is made here.

[0140] In addition, for technical details not described in detail in this embodiment, reference can be made to the method for generating a palmprint image based on a B-spline curve and a CNN provided in any embodiment of the present application, which will not be described in detail here.

[0141] In addition, it should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0142] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0143] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory (ROM) / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of each embodiment of the present application.

[0144] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A palmprint image generation method based on B-spline curve and CNN, characterized in that: include: According to the prior knowledge, the structural features of the palm print are obtained, a square area of ​​a preset size is selected as an ROI area of ​​interest for simulating the palm print according to the structural features, and the ROI area is divided into regions; In different areas after division, parameterized B-spline curves are selected to simulate relevant parameter variables in the palm lines in different areas of ROI; Using a B-spline curve to fit the palm lines according to the ROI area and the relevant parameter variables in the corresponding area to obtain a generated palm line image; Extracting a first image feature from the generated palm line image and a second image feature from the real palm line image using a convolutional neural network (CNN) model; The first image feature and the second image feature are feature-fused to generate a pseudo palmprint image.

2. The palmprint image generation method based on B-spline curve and CNN according to claim 1 is characterized in that: in, The preset size is 7 cm×7 cm, and the ROI area is divided into regions, including: Divide the ROI area into four square areas, wherein each of the divided square areas is uniform and has the same size, and the four square areas are respectively an upper left area, a lower left area, an upper right area, and a lower right area; The upper left area among the four square areas is selected, and the upper left area is also divided into four square areas in the same manner.

3. The palmprint image generation method based on B-spline curve and CNN according to claim 2 is characterized in that: The relevant parameter variables include endpoints and control points, the endpoints include the starting point and the end point of the main lines, and the relevant parameter variables in the palm lines are simulated by using parameterized B-spline curves in different areas of the ROI after division, including: Selecting the starting point of the main pattern from the four square areas divided by the upper left area, and using the starting point as a reference to parameterize the palm crease corresponding to the main pattern using a B-spline curve; Selecting parameter points from the lower left area, the upper right area and the lower right area as the end points of the main lines based on the palm crease; A connecting line is established between the starting point and the end point, and a number of control points are selected from the area formed between the connecting line and the palm crease corresponding to the main pattern.

4. The palmprint image generation method based on B-spline curve and CNN according to claim 3 is characterized in that: The method of using a B-spline curve to fit the palm lines according to the ROI area and the relevant parameter variables in the corresponding area to obtain a generated palm line image includes: The points on the B-spline curve are calculated according to the endpoints and the control points, and the calculation formula is: ； Among them, C i i = (0,1,...,n) is the control point, N i,k (t)(i = 1,2,...,n) is the basis function of the B-spline curve, and the recursive formula is: When k=0, , k=0; When k ≥ 1, , i = 0,1,...,n; Where k is the degree of the B-spline basis function, and U = is the endpoint vector.

5. The palmprint image generation method based on B-spline curve and CNN according to claim 1, characterized in that: The step of fusing the first image feature and the second image feature to generate a pseudo palmprint image includes: Initialize a style image and a content image based on the first image feature and the second image feature; Define a loss function, wherein the loss function includes content loss, style loss, and total variation loss; Selecting a feature extraction part in the CNN model, using five convolutions to output style features based on the style image and the style loss, and using one convolution layer to output content features based on the content image and the content loss; A pseudo palmprint image is generated based on the style feature, the content feature and the total variation loss.

6. The palmprint image generation method based on B-spline curve and CNN according to claim 5, characterized in that: The content loss is used to measure the difference between the generated image and the content image, and the calculation formula is: ； in and are the original content image and the generated image, It is The activation value of the ith filter at position j in the layer, It is The original value of the i-th filter at position j in the layer; The style loss is used to measure the difference between the generated image and the style image, and the calculation formula is: in, and are the original style image and the generated image, is the weight factor of each layer’s contribution to the total loss; The total variation loss is used to reduce the noise in the generated image, and the calculation formula is: ； Among them, α and β are the weight factors for content and style reconstruction, respectively.

7. The palmprint image generation method based on B-spline curve and CNN according to claim 6, characterized in that: expect The calculation formula is: ； in, Represents the number of feature maps, and the size of each feature map is ,in is the height of the feature map multiplied by its width, It is The inner product between the vectorized feature maps i and j in the layer, .

8. A palmprint image generation device based on B-spline curve and CNN, characterized in that: The palmprint image generation device based on B-spline curve and CNN includes: A division module is used to obtain the structural features of the palm print according to prior knowledge, select a square area of ​​a preset size as an interested ROI area of ​​the simulated palm print according to the structural features, and divide the ROI area into regions; A selection module is used to select and use parameterized B-spline curves to simulate relevant parameter variables in the palm lines in different areas of the ROI after division; A fitting module, used for fitting the palm lines using a B-spline curve according to the ROI area and the relevant parameter variables in the corresponding area to obtain a generated palm line image; An extraction module, used for extracting a first image feature from the generated palm line image and a second image feature from the real palm line image using a convolutional neural network (CNN) model; The fusion module is used to fuse the first image feature and the second image feature to generate a pseudo palmprint image.

9. The palmprint image generation device based on B-spline curve and CNN according to claim 8, characterized in that: in, The preset size is 7 cm×7 cm, and the division module is used to divide the ROI area into four square areas, wherein each of the divided square areas is uniform and has the same size, and the four square areas are respectively an upper left area, a lower left area, an upper right area, and a lower right area; The upper left area among the four square areas is selected, and the upper left area is also divided into four square areas in the same manner.

10. A palmprint image generation device based on B-spline curve and CNN, characterized in that: The device comprises: a memory and a processor, and the processor executes the method according to any one of claims 1 to 7 when running a palmprint image generation program based on B-spline curves and CNN stored in the memory.

Images