Vein image quality assessment method, assessment device and terminal

Abstract

The present application provides a method for evaluating the quality of a vein image, which includes: searching for the optimal values ​​of two parameters, brightness and clarity, for evaluating the quality of a vein image based on a neural network training model; inputting a vein image T to be evaluated; scoring the brightness of the vein image T according to the optimal value of the brightness parameter to obtain a brightness score S1, and scoring the clarity of the vein image T according to the optimal value of the clarity parameter to obtain a clarity score S2; scoring the vein feature information of the vein image T to obtain a vein feature score S3; and obtaining a final score evaluation result S of the vein image T by combining the brightness score S1, the clarity score S2, and the vein feature score S3. The present application also provides a method and terminal for evaluating the quality of a vein image. The method, device, and terminal for evaluating the quality of a vein image effectively evaluate the vein image from multiple aspects, which is conducive to obtaining high-quality vein images and reducing the difficulty of subsequent vein image analysis.

Description

Technical Field

[0001] The present application relates to the field of image processing, and more specifically, to a method, an evaluation device and a terminal for evaluating the quality of a vein image. Background Art

[0002] In the prior art, authentication methods based on passwords or instructions are easily stolen or cracked. With the development of technology, authentication systems based on biometrics are becoming more and more popular. Biometrics include face, fingerprint or vein features. The characteristic of human hemoglobin absorbing near-infrared light is used to collect vein feature images under the palm skin. Compared with other biometric features such as fingerprints or irises, palm vein biometrics are located under the human skin, are internal features of the human body and are invisible to the naked eye, so they are difficult to copy and forge; and the palm vein images are collected non-contactly, which is more hygienic and suitable for use in public places.

[0003] During the image acquisition process, the image quality itself needs to be evaluated due to the influence of multiple factors such as the acquisition device itself, whether the imaging process has been compressed, and the illumination and environment during acquisition. This is because image quality often affects the accuracy of subsequent authentication and recognition. Acquiring high-quality images helps reduce the difficulty of subsequent image analysis and helps to improve recognition accuracy.

[0004] Existing image evaluation methods mainly focus on the general quality parameters of images, such as the degree of illumination and clarity of the image, to make judgments. Some existing image quality evaluation methods involve general quality parameters and do not consider the characteristics of the palm vein image itself. For example, when collecting palm vein images, because the thickness of the individual's palm and palm lines affect the extraction of palm vein feature information, it is easy to cause the vein features of the collected palm vein images to be unclear and not rich. In particular, the quality of palm vein images is collected using infrared light, so the brightness of the image is usually much lower than the brightness under natural light. Therefore, it is necessary to set parameters that can collect high-quality illumination quality, and use this illumination quality parameter as the standard to evaluate the subsequent collected vein images, which is conducive to obtaining high-quality vein images. Summary of the invention

[0005] In view of the prior art, the technical problem solved by the present application is a vein image quality assessment method, a vein image quality assessment device and a terminal that can comprehensively evaluate the brightness, clarity and vein feature information of the vein image. The vein image quality assessment method, the vein image quality assessment device and the terminal effectively evaluate the vein image from multiple aspects, which is conducive to obtaining high-quality vein images and reducing the difficulty of subsequent vein image analysis.

[0006] In order to solve the above technical problems, in a first aspect, the present application provides a method for evaluating the quality of a vein image, comprising:

[0007] Searching for optimal values ​​of two parameters, brightness and clarity, for evaluating the quality of vein images based on a neural network training model;

[0008] Input the vein image T to be evaluated;

[0009] Scoring the brightness of the vein image T according to the optimal value of the brightness parameter to obtain a brightness score S1, and scoring the clarity of the vein image T according to the optimal value of the clarity parameter to obtain a clarity score S2;

[0010] Score the vein feature information of the vein image T to obtain a vein feature score S3; and,

[0011] The final score evaluation result S of the vein image T is obtained by integrating the brightness score S1, the clarity score S2 and the vein feature score S3.

[0012] In one embodiment of the application, the step of searching for optimal values ​​of two parameters for evaluating the brightness and clarity of a vein image based on a neural network training model includes:

[0013] Under i kinds of near-infrared light intensities and i kinds of camera parameters for collecting vein images, i vein images are collected i times to obtain i training sets {E 1 ,E 2 ,…,E i},i>1;

[0014] Calculate the i training sets {E 1 ,E 2 ,…, Ei}, and obtain the brightness parameter value set {L 1 ,L 2 ,…,L i} and the clarity parameter value set {R 1 ,R 2 ,…,R i};

[0015] For the training set {E 1 ,E 2 ,…,E i} Perform i-time neural network model training and select the training set E with the smallest training error at the end of the i-time neural network model training j The corresponding brightness parameter value L j and the clarity parameter value R j are the optimal values ​​of brightness parameters L o and the optimal value of the clarity parameter R o ;

[0016] Among them, E j is the jth training set, E j ∈{E 1 ,E 2 ,…,E i}, and 1≤j≤i, L j ∈{L 1 ,L 2 ,…,L i} j ∈{R 1 ,R 2 ,…,R i};L j and R j are the jth training set E j The corresponding brightness parameter value and clarity parameter value.

[0017] In one embodiment of the application, according to the optimal value L of the brightness parameter o The brightness of the vein image T is scored to obtain a brightness score S1, which is calculated as follows:

[0018] According to the optimal value of the clarity parameter R o The clarity of the vein image T is scored to obtain a clarity score S2, which is calculated as follows:

[0019] Among them, L T is the brightness parameter value of the vein image T; R T is the clarity parameter value of the vein image T.

[0020] In one embodiment of the application, the jth training set E is calculated j The corresponding brightness parameter value L j :L j =0.299R j +0.587G j +0.114B j ; Calculate the brightness parameter value L of the vein image T T , L T =0.299R T +0.587G T +0.114B T ; Among them, R j , G j and B j They are training set E j The red, green, and blue channel values ​​of the vein image in R T , G T and GR T are the red, green and blue channel values ​​of the vein image T respectively.

[0021] In one embodiment of the application, the jth training set E is calculated j The corresponding brightness parameter value L j When using the training set E j The red, green and blue channel values ​​of any vein image in the training set E are calculated, or the red, green and blue channel values ​​of any vein image in the training set E are calculated. j The mean of the red, green and blue channels of all vein images is calculated.

[0022] In one embodiment of the application, the jth training set E is calculated j The corresponding clarity parameter value R j Or calculate the clarity parameter value R of the vein image T T The steps include:

[0023] Grayscale the image whose clarity parameter value is to be calculated;

[0024] Perform Laplacian convolution on the grayscale image;

[0025] Calculate the variance of the image after Laplacian convolution;

[0026] The variance value of the image whose clarity parameter value is to be calculated is recorded as the clarity value of the image.

[0027] The image for which the clarity parameter value is to be calculated is the vein image in the training set or the vein image T.

[0028] In one embodiment of the application, each neural network model training is performed based on the CNN model, and the specific steps include:

[0029] Data annotation of the training set, annotating the vein feature information of each vein image in the training set;

[0030] The training set and the annotated vein feature information are input into the CNN training model, and the training error value when the CNN training model converges is recorded.

[0031] In one embodiment of the application, the step of scoring the vein feature information of the vein image T to obtain the vein feature score S3 includes:

[0032] Performing image segmentation on the vein image T to obtain a vein feature information region;

[0033] The vein feature score S3 is obtained by calculating the pixel ratio of the number of pixels occupied by the vein feature information area after image segmentation to the total number of pixels of the vein image T, S3=p 1 / p 2 ;

[0034] Among them, p 1is the number of pixels occupied by the vein feature information area, p 2 is the total number of pixels in the vein image T.

[0035] In one embodiment of the application, the PSPNet method is used to perform image segmentation on the vein image T to obtain vein feature information.

[0036] In one embodiment of the application, a weighted sum is performed on the brightness score S1, the clarity score S2 and the vein feature score S3 to obtain a score evaluation result S of the vein image T, where S=a×S1+b×S2+c×S3; wherein a+b+c=1.

[0037] In the quality assessment method of the vein image, the optimal values ​​of the two parameters of brightness and clarity used to evaluate the quality of the vein image are searched based on the neural network training model, so as to obtain the optimal values ​​of the brightness parameter and the optimal values ​​of the clarity parameter for evaluating the subsequently acquired vein image T; and the brightness and clarity of the vein image T are scored respectively according to the optimal values ​​of the brightness and the clarity parameters; and the vein feature information in the vein image T is scored in combination. The quality assessment method of the vein image effectively and comprehensively evaluates the brightness, clarity and vein feature information of the vein image, and the quality assessment method evaluates the vein image from multiple aspects, which is conducive to obtaining high-quality vein images and reducing the difficulty of subsequent vein image analysis.

[0038] In a second aspect, the present application provides a device for evaluating the quality of a vein image, comprising:

[0039] A vein image acquisition unit, used for acquiring a training set of a neural network model and a vein image T to be scored;

[0040] an optimal value parameter search unit, connected to the vein image acquisition unit, for searching for optimal values ​​of two parameters, brightness and clarity, for evaluating the quality of the vein image based on a neural network training model; and

[0041] an image scoring unit, connected to the vein acquisition unit and the optimal value parameter search unit; used to score the brightness of the vein image T according to the optimal value of brightness searched by the optimal value parameter search unit to obtain a brightness score S1, used to score the clarity of the vein image T according to the optimal value of clarity searched by the optimal value parameter search unit to obtain a clarity score S2, used to score the vein feature information of the vein image T to obtain a vein feature score S3, and used to comprehensively calculate the brightness score S1, the clarity score S2 and the vein feature score S3 to obtain a final score evaluation result S of the vein image T.

[0042] In the vein image quality assessment device, the optimal value parameter search unit searches for the optimal values ​​of the two parameters of brightness and clarity used to assess the quality of the vein image based on the neural network training model; the image scoring unit scores the brightness of the vein image T according to the result of the optimal value of brightness searched by the optimal value parameter search unit, and scores the clarity of the vein image T according to the result of the optimal value of clarity searched by the optimal value parameter search unit, and obtains the vein feature score S3 in combination with the vein feature information scoring of the vein image T. The vein image quality assessment device effectively and comprehensively evaluates the brightness, clarity and vein feature information of the vein image, evaluates the vein image from multiple aspects, is conducive to obtaining high-quality vein images and is conducive to reducing the difficulty of subsequent vein image analysis.

[0043] On the third aspect, the present application also provides a terminal, including a processor, a memory and a camera; the camera is used to capture vein images; the memory is used to store executable codes; the processor is used to execute the executable codes stored in the memory to execute the above-mentioned vein image quality assessment method.

[0044] In the terminal, the optimal values ​​of the two parameters of brightness and clarity used to evaluate the quality of the vein image are searched based on the neural network training model, so as to obtain the optimal values ​​of the brightness parameter and the optimal values ​​of the clarity parameter for evaluating the subsequently acquired vein image T; and the brightness and clarity of the vein image T are scored respectively according to the optimal values ​​of the brightness and the clarity parameters; and the vein feature information in the vein image T is scored in combination. The terminal effectively and comprehensively evaluates the brightness, clarity and vein feature information of the vein image, evaluates the vein image from multiple aspects, is conducive to obtaining high-quality vein images and is conducive to reducing the difficulty of subsequent vein image analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0046] Figure 1 This is a flow chart of a method for evaluating the quality of a vein image according to Embodiment 1 of the present application;

[0047] Figure 2 For the first embodiment of this application Figure 1 A flowchart of the implementation process of step S100 combined with step S300 and step S400;

[0048] Figure 3The calculation of the jth training set E in the first embodiment of the present application is j The corresponding clarity parameter value R j Or calculate the clarity parameter value R of the vein image T T Flow chart of the method;

[0049] Figure 4 This is the quality assessment device for vein images of the second embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with 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.

[0051] It should be noted that when an element is referred to as being "fixed to" or "disposed on" another element, it can be directly on the other element or indirectly on the other element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element.

[0052] It should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0053] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0054] The vein image quality assessment method, vein image quality assessment device and terminal provided in the embodiments of the present application are now described in conjunction with the accompanying drawings.

[0055] Embodiment 1

[0056] Reference Figure 1 The method for evaluating the quality of a vein image provided in the first embodiment of the present application comprises the following steps:

[0057] Step S100: searching for optimal values ​​of two parameters, brightness and clarity, for evaluating vein image quality based on a neural network training model.

[0058] Step S200: Inputting the vein image T to be evaluated.

[0059] Step S300: Score the brightness of the vein image T according to the optimal value of the brightness parameter to obtain a brightness score S1, that is, score the brightness of the vein image T in step S200 according to the optimal value of the brightness parameter searched in step S100 to obtain a brightness score S1.

[0060] Step S400: Score the clarity of the vein image T according to the optimal value of the clarity parameter to obtain a clarity score S2; that is, score the brightness of the vein image T in step S200 according to the optimal value of the clarity parameter searched in step S100 to obtain a brightness score S1.

[0061] Step S500: Score the vein feature information of the vein image T to obtain a vein feature score S3.

[0062] as well as

[0063] Step S600: The final score evaluation result S of the vein image T is obtained by integrating the brightness score S1, the clarity score S2 and the vein feature score S3, that is, the scoring results of step S300, step S400 and step S500 or the final score evaluation result S.

[0064] Reference Figure 2 The search for the optimal values ​​of the two parameters of brightness and clarity for evaluating the quality of vein images based on the neural network training model includes the following steps, that is, the step S100 specifically includes the following steps:

[0065] Step S110: Under the conditions of i kinds of near-infrared light intensities and i kinds of camera parameters for collecting vein images, i vein images are collected to obtain i training sets {E 1 ,E 2 ,…,E i},i>1;

[0066] Step S120: Calculate the i training sets {E 1 ,E 2 ,…,E i}, and obtain the brightness parameter value set {L 1 ,L 2 ,…,L i} and the clarity parameter value set {R 1 ,R2 ,…,R i};

[0067] Step S130: For the training set {E 1 ,E 2 ,…,E i} Perform i-time neural network model training and select the training set E with the smallest training error at the end of the i-time neural network model training j The corresponding brightness parameter value L j and the clarity parameter value R j are the optimal values ​​of brightness parameters L o and the optimal value of the clarity parameter R o ;

[0068] Among them, E j is the jth training set, E j ∈{E 1 ,E 2 ,…,E i}, and 1≤j≤i, L j ∈{L 1 ,L 2 ,…,L i} j ∈{R 1 ,R 2 ,…,R i};L j and R j are the jth training set E j The corresponding brightness parameter value and clarity parameter value. j It can be understood as i {E 1 ,E 2 ,…,E i Any one in the training set.

[0069] Understandably, 1 ,E 2 ,…,E i Any training set E in j The images are all vein images acquired under the same near-infrared light intensity and the same camera parameters.

[0070] It can be understood that the intensity of the near-infrared light irradiating the organism whose vein image is to be collected and the parameters of the camera collecting the vein image will affect the collection result of the vein image. The optimal value of the brightness parameter and the optimal value of the clarity parameter of the vein image with the best quality collected can be searched by step S100. The optimal value of the brightness parameter and the optimal value of the clarity parameter can be used to evaluate the quality of the subsequent vein image collection. The optimal value of the brightness parameter and the optimal value of the clarity parameter can be understood as the corresponding image brightness value and image clarity value when the vein image quality is very good. The camera parameters may include but are not limited to parameters such as sensitivity and focus area.

[0071] In this embodiment, the optimal value of the brightness parameter L can be recorded. o and the optimal value of the clarity parameter R o Corresponding training set E j The intensity of the near-infrared light and the parameters of the camera used during acquisition can be used to acquire the vein image T. At this time, subsequent steps S300, S400, S500 and S600 can verify whether the acquired vein image T is a high-quality image.

[0072] It can be understood that step S130 can be understood as: according to the optimal value L of the brightness parameter o The brightness of the vein image T is scored to obtain a brightness score S1, which is calculated as follows: Step S400 can be understood as: according to the optimal value L of the brightness parameter o The brightness of the vein image T is scored to obtain a brightness score S2, which is calculated as follows: Among them, L T is the brightness parameter value of the vein image T; R T is the clarity parameter value of the vein image T.

[0073] In the above steps S110 to S130, the jth training set E is calculated j The corresponding brightness parameter value L j :L j =0.299R j +0.587G j +0.114B j ; Calculate the brightness parameter value L of the vein image T T , L T =0.299R T +0.587G T +0.114B T ; Among them, R j , G j and B j They are training set E jThe red, green, and blue channel values ​​of the vein image in R T , G T and GR T are the red, green and blue channel values ​​of the vein image T respectively.

[0074] Calculate the jth training set E j The corresponding brightness parameter value L j When using the training set E j The red, green and blue channel values ​​of any vein image in the training set E are calculated, or the red, green and blue channel values ​​of any vein image in the training set E are calculated. j The mean of the red, green and blue channels of all vein images is calculated.

[0075] Reference Figure 3 In the above steps S110 to S130, the jth training set E is calculated j The corresponding clarity parameter value R j Or calculate the clarity parameter value R of the vein image T T The steps include:

[0076] Step S111: graying the image whose clarity parameter value is to be calculated;

[0077] Step S112: performing Laplace convolution on the grayscale processed image;

[0078] Step S113: calculating the variance of the image after Laplace convolution;

[0079] Step S114: record the variance value of the image whose clarity parameter value is to be calculated as the clarity value of the image.

[0080] The image for which the clarity parameter value is to be calculated is the vein image in the training set or the vein image T.

[0081] In step S130, each neural network model training is performed based on the CNN model, and the specific steps include:

[0082] Data annotation of the training set, annotating the vein feature information of each vein image in the training set;

[0083] The training set and the annotated vein feature information are input into the CNN training model, and the training error value when the CNN training model converges is recorded.

[0084] Among them, each CNN training process includes:

[0085] CNN forward propagation: the stage where data propagates from low-level to high-level. During the forward propagation process, the input data passes through the convolution layer and the downsampling layer, extracts the feature vector, and passes the feature vector to the fully connected layer to obtain the output result of the CNN network.

[0086] Calculation of network loss: The error between the network output value and the target value is calculated through the loss function.

[0087] Back propagation of CNN: The stage of training to propagate errors from high levels to low levels. When the result output by CNN does not match the target value, the error calculated by the loss function is used for back propagation. The error is returned layer by layer, the error of each layer of CNN is calculated, and then the weights are updated.

[0088] End of CNN training: When the error calculated by the loss function is equal to or less than the expected value, the model converges and CNN training ends.

[0089] In this embodiment, scoring the vein feature information of the vein image T to obtain the vein feature score S3 specifically includes the following steps, that is, the step S500 specifically includes:

[0090] Step S510: performing image segmentation on the vein image T to obtain a vein feature information region;

[0091] Step S520: Obtain the vein feature score S3 by calculating the pixel ratio of the number of pixels occupied by the vein feature information area after image segmentation to the total number of pixels of the vein image T, S3=p 1 / p 2 ;

[0092] Among them, p 1 is the number of pixels occupied by the vein feature information area, p 2 is the total number of pixels in the vein image T.

[0093] Specifically, the PSPNet method is used to perform image segmentation on the vein image T to obtain vein feature information.

[0094] In step S600, a weighted sum is performed on the brightness score S1, the clarity score S2 and the vein feature score S3 to obtain a score evaluation result S of the vein image T, where S=a×S1+b×S2+c×S3; wherein a+b+c=1.

[0095] In summary, in the quality assessment method of the vein image, the optimal values ​​of the two parameters of brightness and clarity used to evaluate the quality of the vein image are searched based on the neural network training model, so as to obtain the optimal values ​​of the brightness parameter and the optimal values ​​of the clarity parameter for evaluating the subsequent acquisition of the vein image T; and the brightness and clarity of the vein image T are scored respectively according to the optimal values ​​of the brightness and the clarity parameters; and the vein feature information in the vein image T is scored in combination. The quality assessment method of the vein image effectively and comprehensively evaluates the brightness, clarity and vein feature information of the vein image. The quality assessment method evaluates the vein image from multiple aspects, which is conducive to obtaining high-quality vein images and reducing the difficulty of subsequent vein image analysis.

[0096] Embodiment 2

[0097] Reference Figure 3 The vein image quality assessment device 100 provided in the second embodiment of the present application includes a vein image acquisition unit 101, an optimal value parameter search unit 102 and an image scoring unit 103.

[0098] Specifically, the vein image acquisition unit 101 is used to acquire the training set of the neural network model and the vein image T to be scored. The optimal value parameter search unit 102 is connected to the vein image acquisition unit 101, and is used to search for the optimal values ​​of the two parameters of brightness and clarity for evaluating the quality of the vein image based on the neural network training model. The image scoring unit 103 is connected to the vein acquisition unit and the optimal value parameter search unit 102; it is used to score the brightness of the vein image T according to the optimal value of brightness searched by the optimal value parameter search unit 102 to obtain a brightness score S1, to score the clarity of the vein image T according to the optimal value of clarity searched by the optimal value parameter search unit 102 to obtain a clarity score S2, to score the vein feature information of the vein image T to obtain a vein feature score S3, and to obtain a final score evaluation result S of the vein image T by comprehensively combining the brightness score S1, the clarity score S2 and the vein feature score S3.

[0099] In this embodiment, the vein image acquisition unit 101 includes a camera, a plurality of near-infrared LEDs emitting near-infrared light, and an infrared image sensor. The optimal value parameter search unit 102 and the image scoring unit 103 may be, but are not limited to, embedded chips.

[0100] In the vein image quality assessment device 100, the optimal value parameter search unit 102 searches for the optimal values ​​of the two parameters of brightness and clarity for evaluating the quality of the vein image based on the neural network training model; the image scoring unit 103 scores the brightness of the vein image T according to the result of the optimal value of the brightness searched by the optimal value parameter search unit 102, and scores the clarity of the vein image T according to the result of the optimal value of the clarity searched by the optimal value parameter search unit 102, and obtains the vein feature score S3 in combination with the vein feature information scoring of the vein image T. The vein image quality assessment device 100 effectively and comprehensively evaluates the brightness, clarity and vein feature information of the vein image, evaluates the vein image from multiple aspects, is conducive to obtaining high-quality vein images and is conducive to reducing the difficulty of subsequent vein image analysis.

[0101] Embodiment 3

[0102] Embodiment 3 of the present application provides a terminal including a processor, a memory and a camera; the camera is used to capture vein images; the memory is used to store executable codes; the processor is used to execute the executable codes stored in the memory to execute the above-mentioned vein image quality assessment method.

[0103] In the terminal, the optimal values ​​of the two parameters of brightness and clarity used to evaluate the quality of the vein image are searched based on the neural network training model, so as to obtain the optimal values ​​of the brightness parameter and the optimal values ​​of the clarity parameter for evaluating the subsequently acquired vein image T; and the brightness and clarity of the vein image T are scored respectively according to the optimal values ​​of the brightness and the clarity parameters; and the vein feature information in the vein image T is scored in combination. The terminal effectively and comprehensively evaluates the brightness, clarity and vein feature information of the vein image, evaluates the vein image from multiple aspects, is conducive to obtaining high-quality vein images and is conducive to reducing the difficulty of subsequent vein image analysis.

[0104] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for evaluating the quality of a vein image, characterized in that: include: Searching for optimal values ​​of two parameters, brightness and clarity, for evaluating the quality of vein images based on a neural network training model; Input the vein image T to be evaluated; The brightness of the vein image T is scored according to the optimal value of the brightness parameter to obtain a brightness score S1, and the clarity of the vein image T is scored according to the optimal value of the clarity parameter to obtain a clarity score S2; according to the optimal value of the brightness parameter L o The brightness of the vein image T is scored to obtain a brightness score S1, which is calculated as follows: According to the optimal value of the clarity parameter R o The clarity of the vein image T is scored to obtain a clarity score S2, which is calculated as follows: Among them, L T is the brightness parameter value of the vein image T; R T is the clarity parameter value of the vein image T; scoring the vein feature information of the vein image T to obtain a vein feature score S3; and The final score evaluation result S of the vein image T is obtained by integrating the brightness score S1, the clarity score S2 and the vein feature score S3; The step of searching for optimal values ​​of two parameters for evaluating the brightness and clarity of the vein image based on the neural network training model includes: Under the conditions of i kinds of near-infrared light intensities and i kinds of camera parameters for collecting vein images, i vein images are collected to obtain i training sets {E1, E2, …, E i },i>1; Calculate the i training sets {E 1, E2,…,E i }, and obtain the brightness parameter value set {L1, L2, …, L i } and the clarity parameter value set {R1, R2, …, R i }; For the training set {E 1, E2,…,E i } Perform i-time neural network model training and select the training set E with the smallest training error at the end of the i-time neural network model training j The corresponding brightness parameter value L j and the clarity parameter value R j are the optimal values ​​of brightness parameter L o and the optimal value of clarity parameter R o ; Each neural network model training is based on the CNN model. The specific steps include: Data annotation of the training set, annotating the vein feature information of each vein image in the training set; Input the training set and the annotated vein feature information into a CNN training model, wherein the CNN training model is used for palm vein recognition and authentication, and record the training error value when the CNN training model converges; Among them, E j is the jth training set, E j ∈{E1,E2,…,E i }, and 1≤j≤i, L j ∈{L1,L2,…,L i } j ∈{R1,R2,…,R i };L j and R j are the jth training set E j The corresponding brightness parameter value and clarity parameter value; Calculate the jth training set E j The corresponding clarity parameter value R j Or calculate the clarity parameter value R of the vein image T T The steps include: Grayscale the image whose clarity parameter value is to be calculated; Perform Laplacian convolution on the grayscale image; Calculate the variance of the image after Laplacian convolution; The variance value of the image whose clarity parameter value is to be calculated is recorded as the clarity value of the image; The image whose clarity parameter value is to be calculated is the vein image in the training set or the vein image T; The brightness score S1, the clarity score S2 and the vein feature score S3 are weightedly summed to obtain the score evaluation result S of the vein image T, S=a×S1+b×S2+c×S3; wherein a+b+c=1.

2. The method for evaluating the quality of a vein image according to claim 1, wherein: Calculate the jth training set E j The corresponding brightness parameter value L j :L j =0.299R j +0.587G j +0.114B j ; Calculate the brightness parameter value L of the vein image T T , L T＝ 0.299R T +0.587G T +0.114B T ; Among them, R j , G j and B j They are training set E j The red, green, and blue channel values ​​of the vein image in R T , G T and GR T are the red, green and blue channel values ​​of the vein image T respectively.

3. The method for evaluating the quality of a vein image according to claim 2, wherein: Calculate the jth training set E j The corresponding brightness parameter value L j When using the training set E j The red, green and blue channel values ​​of any vein image in the training set E are calculated, or the red, green and blue channel values ​​of any vein image in the training set E are calculated. j The mean of the red, green and blue channels of all vein images is calculated.

4. The method for evaluating the quality of a vein image according to claim 1, wherein: The step of scoring the vein feature information of the vein image T to obtain a vein feature score S3 comprises: Performing image segmentation on the vein image T to obtain a vein feature information region; The vein feature score S3 is obtained by calculating the pixel ratio of the number of pixels occupied by the vein feature information area after image segmentation to the total number of pixels of the vein image T, where S3 = p1 / p2; Among them, p1 is the number of pixels occupied by the vein feature information area, and p2 is the total number of pixels in the vein image T.

5. The method for evaluating the quality of a vein image according to claim 4, wherein: The PSPNet method is used to segment the vein image T to obtain vein feature information.

6. A terminal, characterized in that: It comprises a processor, a memory and a camera; the camera is used to capture vein images; the memory is used to store executable codes; the processor is used to execute the executable codes stored in the memory to execute the vein image quality assessment method as described in any one of claims 1 to 5.

Images