TW202111498A - Fingerprint recognition method, chip and electronic device - Google Patents

Abstract

The present invention provides a fingerprint recognition method, the method includes: acquiring an image to be recognized collected by a fingerprint collection device, and dividing the image into a plurality of sub-images; extracting feature values of each sub-image; inputting the feature values of each sub-image into a preset machine learning model, and determining whether the sub-image is a finger image respectively; determining the image as a finger image if the number of the sub-image determined as the finger image exceeds a preset value; and recognizing a fingerprint in the image. The invention also provides a fingerprint recognition chip and an electronic device.

Description

Fingerprint identification method, chip and electronic device

The invention relates to the technical field of image recognition, in particular to a fingerprint recognition method, a chip and an electronic device.

Due to the convenience and security of biometrics, biometrics technology has broad application prospects in the fields of identity authentication and network security. Fingerprint recognition technology, as one of the biometric identification technologies, is applied to various terminal devices such as smart phones to realize identity authentication functions such as mobile phone unlocking and online payment. However, when fingerprint recognition technology is applied to terminal devices such as smart phones, it may happen that other human body parts or other items other than fingers touch the fingerprint collection device by mistake, such as clothing pockets, smooth skin on non-finger areas, keys Other objects such as, peels, etc. may also contact the fingerprint collection device for various reasons. If these non-finger objects are also directly identified and matched with fingerprints, it wastes computing resources and is not conducive to the improvement of fingerprint recognition efficiency, and even non-finger images may appear. Like the case of misidentification as a fingerprint.

In view of the above problems, the present invention proposes a fingerprint identification method, chip and electronic device to improve the efficiency and accuracy of fingerprint detection.

The first aspect of the present application provides a fingerprint identification method, which includes: Acquiring the image to be recognized collected by the fingerprint acquisition device, and dividing the image into a plurality of sub-images; Extracting the feature value of each sub-image; Input the feature value of each sub-image to a preset machine learning model, and determine whether each sub-image is a finger image; Obtain the judgment result of the sub-images, and when the sub-images exceeding the preset number are judged to be finger images, it is determined that the image to be recognized is a finger image; and Perform fingerprint recognition on the finger image.

Preferably, the extracting the feature value of each sub-image includes: Generating a normalized grayscale image according to each of the sub-images; Generating a normalized gradient image of each of the sub-images; Generating a normalized gray-level gradient co-occurrence matrix according to the normalized gray-level image and the normalized gradient image; Extracting the feature value of each sub-image based on the normalized gray gradient co-occurrence matrix.

Preferably, the feature value includes any one or more of the following: small gradient advantage, large gradient advantage, unevenness of grayscale distribution, unevenness of gradient distribution, energy, grayscale average, gradient average, grayscale mean square error, gradient Mean square error, correlation, gray entropy, gradient entropy, mixed entropy, difference moment, inverse difference moment.

Optionally, the feature value of each sub-image can also be extracted based on the gray level co-occurrence matrix.

Preferably, the preset machine learning model includes any one or more of a neural network model, a support vector machine model, a classification model based on a decision tree, and a Bayesian classification model.

Preferably, the training method of the machine learning model includes: Acquire a sample data set, the sample data set includes a positive sample and a negative sample, the positive sample is the feature value corresponding to the finger image, the corresponding label is the first label, and the negative sample is the feature corresponding to the non-finger image Value, the corresponding label is the second label; Dividing the sample data set into a training set and a test set; Training the machine learning model by using the training set; Use the test set to test the trained machine learning model, and adjust the parameters of the trained machine learning model according to the test results.

Preferably, after determining that the image to be recognized is a finger image, the fingerprint recognition method further detects the effective area of the fingerprint in the finger image, including: Filtering the image to be recognized to obtain a filtered grayscale image; Calculating the gradient of each pixel in the filtered gray image to obtain a gradient direction image; Dividing the gradient direction image into multiple sub-blocks, and calculating the variance of each sub-block; The variance of each sub-block is compared with a set threshold. If the variance of the sub-block is greater than the threshold, the sub-block is determined to be a valid area, and the flag bit of the sub-block is set to the first A value; otherwise, the sub-block is determined to be an invalid area, and the flag bit of the sub-block is set to a second value.

Preferably, the fingerprint identification method further includes: for the image sub-blocks determined to be invalid regions, using a four-neighbors judgment method or an eight-neighbors judgment method, according to the value of the flag bits of the adjacent areas of the invalid area Determine the value of the invalid area flag bit.

The second aspect of the present application provides a fingerprint recognition chip, the chip including: The image segmentation module is used to obtain the fingerprint image to be recognized, and divide the fingerprint image into a plurality of sub-images; The feature extraction module is used to extract the feature value of each sub-image shown; The input module is used for inputting the feature value of each sub-image to a preset machine learning model, and separately determining whether each sub-image is a finger image; and The judging module is used for judging the result of the sub-image. When the sub-images exceeding the preset number are judged to be a finger, the fingerprint image is determined to be a finger image.

A third aspect of the present invention provides an electronic device, the electronic device comprising: a fingerprint collection unit for collecting fingerprint images; a processor; and a memory in which a plurality of program modules are stored, and the plurality of program modules are stored in the memory; Each program module is loaded by the processor and executes the aforementioned fingerprint identification method.

The fingerprint identification method, chip and electronic device of the present invention can automatically identify non-finger regions by segmenting the image to be identified and extracting feature values, combined with a machine learning model, and can prevent attacks from other living parts or objects except fingers , Reduce the misjudgment rate of fingerprint recognition.

In order to be able to understand the above objectives, features and advantages of the present invention more clearly, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments of the application and the features in the embodiments can be combined with each other if there is no conflict.

In the following description, many specific details are set forth in order to fully understand the present invention, and the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field of the present invention. The terms used in the description of the present invention herein are only for the purpose of describing specific embodiments, and are not intended to limit the present invention.

Please refer to FIG. 1. FIG. 1 is a schematic flowchart of a fingerprint identification method according to an embodiment of the present invention. According to different needs, the order of the steps in the flowchart can be changed, and some steps can be omitted. For ease of description, only the parts related to the embodiment of the present invention are shown.

As shown in Figure 1, the fingerprint identification method includes the following steps.

Step S1: Obtain the image to be recognized collected by the fingerprint collecting device, and divide the image into a plurality of sub-images.

The fingerprint collection device may be installed in smart terminals such as mobile phones, tablet computers, industrial equipment, etc., and is used to collect user fingerprints for identity authentication. The fingerprint collection device may also be installed in the attendance device for collecting user fingerprints for attendance and the like. The fingerprint collection device may collect fingerprint images by means such as optical fingerprint collection technology, capacitive sensor fingerprint collection technology, ultrasonic fingerprint collection technology, or electromagnetic wave fingerprint collection technology.

In the embodiment of the present invention, the image to be recognized is a grayscale image. In the existing fingerprint collection equipment, some of the collected images are grayscale images, and some of the collected images are not grayscale images. When the collected images are not grayscale images, the step S1 also includes converting the image collected by the fingerprint collection device into a grayscale image.

After the image to be recognized is acquired, the image is equally divided into m*n sub-images that do not overlap each other, where m and n are both positive integers greater than 1, and the values of m and n can be The same or different. For example, an acquired image may be equally divided into four sub-images in a 2*2 manner, or the image may be equally divided into six sub-images in a 2*3 manner.

Step S2: Extract the feature value of each sub-image.

In the first embodiment of the present invention, the feature value of each sub-image is extracted by calculating the gray gradient co-occurrence matrix of each sub-image, as shown in FIG. The feature value of a sub-image specifically includes the following steps:

Step S201: Generate a normalized grayscale image of each sub-image.

Wherein, generating the normalized gray-scale image can be achieved by using a mean-variance normalization algorithm or a gray-scale transformation normalization algorithm.

Mean variance normalization is to convert the images collected at different times and different brightness to the same gray-scale mean and variance standard image; gray-scale transformation normalization is to use the gray-scale stretching method to convert the original image The grayscale distribution of is extended to an image with the entire grayscale. Both algorithms are existing algorithms and will not be described in detail here.

Step S202: Generate a normalized gradient image of each sub-image.

In one embodiment, the normalized gradient image can be generated using the Sobel operator. The Sobel operator is a discrete first-order difference operator used to calculate the approximate value of the first-order gradient of the image brightness function, using the Sobel operator Element to obtain the gradient direction image G(x,y) of the grayscale image f(x,y), where x represents the abscissa of the pixel point in the grayscale image, and y represents the grayscale image The ordinate of the center primitive point.

。The calculation formula of the gradient direction image G(x, y) is as follows:

.

Wherein, Gx and Gy are the gradient amplitudes of the original grayscale image on the x-axis and y-axis, respectively, which are calculated as follows: Gx={f(x+1,y-1)+2f(x+1,y)+f(x+1,y+1)}-{f(x-1,y-1)+2f(x -1,y)+f(x-1,y+1)}; Gy={f(x-1,y+1)+2f(x,y+1)+f(x+1,y+1)}-{f(x-1,y-1)+2f(x ,y-1)+f(x+1,y-1)};

In other embodiments, the gradient operand may also be one of Roberts operand, Prewitt operand, or Laplace operand.

Step S203: Generate a normalized gray gradient co-occurrence matrix according to the normalized gray image and the normalized gradient image.

定義為歸一化灰度圖像

及歸一化梯度圖像

中具有灰度值

和梯度值

的圖元數。即

； 其中，

為歸一化灰度圖像中的值，

為歸一化梯度圖像中的值，

為歸一化灰度圖像量級，

為歸一化梯度圖像量級。Gray-GradientCo-occurrence Matrix (GGCM) is a matrix composed of gray level nGray and gradient level nGrad. Elements of gray gradient co-occurrence matrix

Defined as a normalized grayscale image

Normalized gradient image

Gray value

And gradient value

The number of primitives. which is

; among them,

Is the value in the normalized grayscale image,

Is the value in the normalized gradient image,

Is the normalized gray image magnitude,

Is the normalized gradient image magnitude.

為歸一化灰度梯度共生矩陣，有

。Assume

Is the normalized gray gradient co-occurrence matrix, with

.

Step S204: Extract the feature value of each sub-image based on the gray gradient co-occurrence matrix.

Wherein, the eigenvalues extracted by the gray gradient co-occurrence matrix include one or more of the following: small gradient advantage, large gradient advantage, unevenness of grayscale distribution, unevenness of gradient distribution, energy, grayscale average, gradient average, Gray-level mean square error, gradient mean square error, correlation, gray-level entropy, gradient entropy, mixed entropy, difference moment, inverse difference moment.

； 2) 大梯度優勢計算方法如下：

； 3) 灰度分佈不均勻性計算方法如下：

； 4) 梯度分佈不均勻性計算方法如下：

； 5) 能量計算方法如下：

； 6) 灰度平均計算方法如下：

； 7) 梯度平均計算方法如下：

； 8) 灰度均方差計算方法如下：

； 9) 梯度均方差計算方法如下：

； 10) 相關性計算方法如下：

； 11) 灰度熵計算方法如下：

； 12) 梯度熵計算方法如下：

； 13) 混合熵計算方法如下：

； 14) 差分矩計算方法如下：

； 15) 逆差分矩計算方法如下：

。Among them, 1) The calculation method of small gradient advantage is as follows:

; 2) The calculation method of large gradient advantage is as follows:

; 3) The calculation method for the unevenness of gray distribution is as follows:

; 4) The calculation method of gradient distribution inhomogeneity is as follows:

; 5) The energy calculation method is as follows:

; 6) The gray average calculation method is as follows:

; 7) The gradient average calculation method is as follows:

8) The calculation method of the gray mean square error is as follows:

9) The calculation method of gradient mean square error is as follows:

10) The correlation calculation method is as follows:

11) The gray entropy calculation method is as follows:

; 12) The gradient entropy calculation method is as follows:

13) The calculation method of mixed entropy is as follows:

; 14) The calculation method of the difference moment is as follows:

15) The calculation method of the inverse difference moment is as follows:

.

It can be understood that in the above steps, step S201 and step S202 are in no particular order. The gray gradient co-occurrence matrix can extract rich fingerprint information in the finger image, which helps to improve the accuracy of finger recognition. Because the gray scale is the basis of an image, the gradient is the element that constitutes the edge contour of the image, and the main information of the image is provided by the edge contour of the image. The gray-level gradient co-occurrence matrix can clearly describe the distribution of the gray level and gradient of each pixel point in the image, and it also gives the spatial relationship between each image point and its domain image point, which can have a good effect on the texture of the image. Delineation can be reflected in the direction of the gradient for the directional texture.

In the second embodiment of the present invention, the eigenvalue of each sub-image can also be extracted by calculating the gray-level co-occurrence matrix of each sub-image.

Gray-level co-occurrence matrix is a method to describe texture by studying the spatial correlation characteristics of image gray. The gray level co-occurrence matrix is an L*L matrix, where L is the number of gray levels of the image after quantization. The dimension of the gray level co-occurrence matrix has nothing to do with the image to be processed (gray image or converted gray image), but is determined by the number of gray levels L of the gray image, assuming the image to be processed The size is a*b, and the number of gray levels is L, then the size of the gray-level co-occurrence matrix is L*L. The larger the image gray level L, the clearer the image, and the more truly reflecting the original image itself; however, the larger the L, the greater the dimension of the gray-level co-occurrence matrix, which greatly increases the amount of calculation. Generally speaking, the gray value range of the gray image is 0-255, and the value range of the gray level is [8,16,32,64,128,256]. When calculating the gray level co-occurrence matrix and its eigenvalues, in order to reduce the amount of calculation, it is more preferable to set the gray level to 8 or 16.

。In the second embodiment, the gray level L=8 is preferably used. The following methods can be used to quantize the gray level with a value range of 0-255 to a total of 8 quantization levels from 0-7:

.

Among them, g represents the original gray value, and Ng represents the quantized gray level.

It can be understood that in other embodiments, the gray value in the image can also be quantized into gray levels of other values. The present invention does not limit this, and the embodiments of each quantization level should fall within the protection scope of the present invention. Inside.

In the second embodiment, the values of elements in the gray-level co-occurrence matrix are selected as follows: Count the number of point pairs (two pixel points are called a point pair) in the grayscale image that meet the following conditions p: the distance between the two pixel points in the point pair is d (d is also called step length or displacement ), the direction is θ (that is, the angle between the line of two points and the horizontal axis), and the quantized gray levels are i and j respectively. Then p is the element value of the i-th row and j-th column in the gray-level co-occurrence matrix. Among them, the displacement d and the direction θ need to be initialized and remain unchanged during the calculation process.

In the second embodiment shown, it is preferable that the displacement d=1, that is, two adjacent graphic element points are regarded as a point pair. The value of the direction θ is 0 degrees, 45 degrees, 90 degrees, and 135 degrees. That is to say, in this embodiment, each element (i, j) in the gray-level co-occurrence matrix represents the number of times the gray level i and the gray level j are adjacent in the direction θ in the image.

In one embodiment, the eigenvalues can be extracted through a gray-level co-occurrence matrix in one direction. In another embodiment, the gray-level co-occurrence matrix is obtained by averaging multiple gray-level co-occurrence matrices corresponding to different directions. For example, the gray-level co-occurrence matrix of the sub-image passes 0 degrees, 45 degrees, and 90 degrees. The gray-level co-occurrence matrix corresponding to the four directions of degrees and 135 degrees is averaged, that is, the gray-level co-occurrence matrix GLCM is:

In the second embodiment, the feature values of the sub-image extracted based on the gray-level co-occurrence matrix include contrast (Contrast), correlation (Correlation), homogeneity (Homogeneity), and maximum probability (Maximun Probalility) .

。Among them, the calculation formula of contrast is:

.

，其中，

。The calculation formula for correlation is:

,among them,

.

。 The calculation formula for uniformity is:

.

。The formula for calculating the maximum probability is:

.

In practical applications, 14 eigenvalues can generally be extracted from the gray-level co-occurrence matrix. Therefore, it can be understood that in other embodiments, other eigenvalues can also be extracted based on the gray-level co-occurrence matrix as the sub-image. The characteristic values should all fall within the protection scope of the present invention.

In other embodiments of the present invention, extracting the feature value of the sub-image can also be implemented by other feature extraction methods, such as extracting the LBP (Local Binary Pattern) feature of the image.

Step S3: Input the feature value of each sub-image into a preset machine learning model, and determine whether each sub-image is a finger image.

Wherein, the input of the machine learning model is the feature value of the image. In the first embodiment, the implementation feature value is the feature value extracted from the gray gradient co-occurrence matrix. In the second embodiment, the feature value includes Contrast, Correlation, Homogeneity, and Maximun Probalility extracted based on the gray level co-occurrence matrix. The output of the machine learning model is the classification result of the image, that is, whether it is a finger image.

In the embodiment of the present invention, the machine learning model may be, but is not limited to, a BP neural network model, a support vector machine model, a classification model based on a decision tree, a Bayesian classification model, and the like. The present invention does not enumerate the machine learning models one by one, and all relevant modified embodiments should fall within the protection scope of the present invention.

As shown in FIG. 3, it is a schematic diagram of the process flow of a neural network model training method in an embodiment of the present invention. In this embodiment, the training method of the neural network model includes the following steps.

Step S301: Obtain a sample data set, the sample data set includes positive samples and negative samples, the positive samples are the feature values extracted based on the gray level co-occurrence matrix corresponding to the finger image, and the corresponding label is 1, the negative sample It is the feature value extracted based on the gray-level co-occurrence matrix corresponding to the non-finger image, and the corresponding label is 0. For example, non-finger pictures such as peels, conductive foam, etc., have a label of 0.

In this embodiment, the positive and negative sample image feature values include contrast (Contrast), correlation (Correlation), homogeneity (Homogeneity), and maximum probability (Maximun Probalility) extracted based on the gray level co-occurrence matrix.

Step S302: Divide the sample data set into a training set and a test set.

For example, for the above positive and negative sample images, 70% of them are classified as a training set, and the purpose is to train the neural network model, and the remaining 30% are classified into another category as a test set , Its purpose is to test the classification performance of the neural network model.

Step S303: Use the training set to train the neural network model.

The process of inputting the training set into the established neural network model for model training can be implemented by means in the prior art, and will not be described in detail here. In some embodiments, using the training set to train the neural network model may further include: deploying the training of the deep neural network model on multiple graphics processing units (GPUs) for distributed training. For example, through Tensorflow's decentralized training principle, the training of the model can be deployed on multiple graphics processors for decentralized training, which can shorten the training time of the model and accelerate the convergence of the model.

Step S304: Use the test set to test the trained neural network model, and adjust the parameters of the trained neural network model according to the test results.

In this embodiment, using the test set to test the trained neural network model may include: sending the sample data in the test set to the trained neural network model, and the neural network model will test the neural network model. The feature value of the sample image is recognized, that is, it is judged whether the sample image is a finger image, and the result is compared with its own label to obtain the classification accuracy of the neural network model. It is determined whether the training needs to be continued according to the correctness rate, and if the correctness rate reaches the preset value, the model training is completed.

Wherein, the training of the machine learning model can be done offline. After the training of the neural network model is completed, input the sub-image to be recognized into the neural network model to identify whether the sub-image is a finger image, and realize the automatic recognition of the image, saving time in the whole process , Labor saving, not only improves the efficiency of finger image recognition, but also greatly improves the accuracy of finger image recognition.

Step S4: Obtain the judgment result of the sub-image, and when the sub-image exceeding the preset number is judged to be a finger, it is determined that the fingerprint image is a finger image.

The preset number can be set as required, and when the number of sub-images determined to be fingers is less than the preset number, it is determined that the fingerprint image is a non-finger image. For example, when the fingerprint image is divided into 4 sub-images, if 3 or more of the 4 sub-images are determined to be fingers, it is determined that the fingerprint image is a finger If only 2 or less of the 4 sub-images are determined to be a finger, it is determined that the fingerprint image is not a finger image.

In some embodiments of the present invention, the sub-images of the image to be recognized may be sequentially input into the machine learning model, when it is determined that more than a preset number of sub-images have been determined to be finger images , The input and discrimination of other sub-images can be stopped, and the image to be recognized is directly determined as a finger image, which can further save the amount of calculation and improve the recognition efficiency of the finger image. For example, when the image to be recognized is divided into 6 sub-images, if the first 4 sub-images are all determined to be finger images, and the preset number has been reached, there is no need to divide the remaining images. The 2 sub-images of is then input into the machine learning model, thereby saving computing time.

The present invention divides the image collected by the fingerprint acquisition device into multiple sub-images for feature extraction, and uses a machine learning model to identify whether it is a finger image, and only performs fingerprint identification and matching authentication when it is identified as a finger image. Excluding non-finger images can effectively avoid attacks by foreign objects such as non-finger items (such as clothing pockets, peels, smooth skin), reduce the amount of calculation for fingerprint recognition, and improve the accuracy of fingerprint recognition.

Further, when it is determined that the image to be recognized is an image of a finger, but not an image of other non-finger objects, fingerprint recognition and matching can be started. However, in the fingerprint matching process, the fingerprint texture of some fingerprint regions is relatively fuzzy, and the fingerprint texture of some fingerprint regions is relatively clear. The fuzzy area of the fingerprint texture is likely to cause fingerprint recognition errors, and the fingerprint recognition and matching operations of the entire finger are also relatively large.

Therefore, in order to further improve the accuracy and efficiency of fingerprint recognition, in this embodiment, after determining that the image to be recognized is a finger image, the fingerprint recognition method preferably further includes step S5: The effective area of the fingerprint is detected. Wherein, the effective area refers to an area with clear fingerprint lines in an image, and an invalid area refers to an area with unclear fingerprint lines in an image. It can be understood that in other embodiments, the step S5 may also be omitted.

Specifically, refer to FIG. 4, which is a schematic flowchart of a method for detecting a valid area of a fingerprint in an embodiment of the present invention.

In this embodiment, the method for detecting the effective area of the fingerprint includes the following steps.

Step S401: Perform filtering processing on the image to be identified collected by the fingerprint collection device to obtain a filtered grayscale image.

The filtering method may be, but is not limited to, any one of median filtering, mean filtering, Gaussian filtering, and bilateral filtering. By filtering the image to be recognized, the influence of noise can be eliminated. In a preferred embodiment, mean filtering is used to perform filtering processing on the image to be recognized.

Step S402: Calculate the gradient of each graphic element in the filtered grayscale image to obtain a gradient direction image.

Specifically, the Sobel operator can be used to obtain the gradient direction image G(x,y) of the grayscale image f(x,y), where x represents the abscissa of the pixel point in the grayscale image, y represents the ordinate of the pixel point in the grayscale image.

。The calculation formula of the gradient direction image G(x, y) is as follows:

.

Wherein, Gx and Gy are the gradient amplitudes of the original grayscale image on the x-axis and y-axis, respectively, which are calculated as follows: Gx={f(x+1,y-1)+2f(x+1,y)+f(x+1,y+1)}-{f(x-1,y-1)+2f(x -1,y)+f(x-1,y+1)}; Gy={f(x-1,y+1)+2f(x,y+1)+f(x+1,y+1)}-{f(x-1,y-1)+2f(x ,y-1)+f(x+1,y-1)};

In other embodiments, other operation elements may also be used to calculate the gradient direction image.

Step S403: Divide the gradient direction image into multiple sub-blocks, and calculate the variance of each sub-block.

The present invention does not specifically limit the segmentation of the gradient direction image. In a preferred embodiment, the gradient direction image can be divided into 16*16 sub-blocks. In other embodiments, it can also be divided into 16*16 sub-blocks. Other number of sub-blocks. The calculation of the variance of each sub-block can be achieved by using formulas in the prior art, which will not be described in detail here.

Step S404: Compare the variance of each sub-block with a set threshold, and if it is greater than the threshold, determine that the sub-block is a valid area, and set the flag bit of the sub-block to the first value; otherwise, It is determined that the sub-block is an invalid area and the flag bit of the sub-block is set to a second value. In this embodiment, the first value is 1, and the second value is 0.

Wherein, the threshold value can be set as required, and the present invention does not specifically limit this.

After detecting the effective area with clear fingerprint lines in the finger image, fingerprint identification and matching can be performed through the effective area, which can improve the accuracy of fingerprint recognition, and the effective area identification and matching are relative to the entire finger image. The amount of calculation for image recognition and comparison is greatly reduced, so the calculation efficiency can also be improved.

In a preferred embodiment of the present invention, in order to further improve the accuracy of fingerprint recognition, after detecting the effective area of the fingerprint, for the sub-blocks determined to be invalid areas, the method further includes: according to the four-neighbor judgment method according to The values of the four adjacent area flag bits on the upper, lower, left, and right of the invalid area re-determine the invalid area.

The four-neighborhood determination method is that when the values of the flag bits of the four neighbors of the area are all of the same type, it means that the area belongs to the category of its neighborhood area, then the area is classified into its neighborhood category, and the flag bit of the area is The value of the element is set to the value of the flag bit of the neighborhood, otherwise, the value of the flag bit of the area remains unchanged. For example, as shown in Figure 5, suppose the position of the invalid region is f (x, y), the value of the flag bit is 0, and it is assumed that the upper adjacent region f (x, y-1) and the lower adjacent region f(x,y+1), the adjacent area on the left f(x-1,y), and the adjacent area on the right f(x+1,y) have the flag bit value of 1, then according to The four neighbors re-determine the value of the flag bit of the invalid area as 1, that is, according to the four-neighbor judgment method, the invalid area is re-determined as a valid area.

或

，則將

置為1，即有效； 2）  無效區域在待識別圖像右上角時的判定方法：若

或

，則將

置為1，即有效； 3）  無效區域在待識別圖像左下角時的判定方法：若

或

，則將

置為1，即有效； 4）  無效區域在待識別圖像右下角時的判定方法：若

或

，則將

置為1，即有效； 5）  無效區域在待識別圖像上邊界時的判定方法：若

或

，則將

置為1，即有效； 6）  無效區域在待識別圖像下邊界時的判定方法：若

或

，則將

置為1，即有效； 7）  無效區域在待識別圖像左邊界時的判定方法：若

或

，則將

置為1，即有效； 8）  無效區域在待識別圖像右邊界時的判定方法：若

或

，則將

置為1，即有效； 9）  無效區域在待識別圖像中心區域時的判定方法：若

，則將

置為1，即有效。When re-judging the invalid area in the image according to the four-neighborhood method, due to the difference in the position of the invalid area in the image, some image blocks may not have enough neighborhoods of four. For example, the invalid area is in the upper left corner of the original image. In the block, there may only be two neighborhoods on the right and the bottom. As shown in Figure 6, the position of the invalid area in the image may be in the upper left corner, upper right corner, lower left corner, lower right corner, and the upper, lower, left, and right boundaries. For these special cases, the location of the invalid area may be determined according to the four neighborhoods. When re-determining the value of the invalid area, the judgment is made according to the following methods: 1) The judgment method when the invalid area is in the upper left corner of the image to be recognized: if

or

, Then

Set to 1, that is valid; 2) Judgment method when the invalid area is in the upper right corner of the image to be recognized: If

or

, Then

Set to 1, that is valid; 3) Judgment method when the invalid area is in the lower left corner of the image to be recognized: If

or

, Then

Set to 1, that is valid; 4) Judgment method when the invalid area is in the lower right corner of the image to be recognized: If

or

, Then

Set to 1, that is valid; 5) Judgment method when the invalid area is on the upper boundary of the image to be recognized: If

or

, Then

Set to 1, that is valid; 6) Judgment method when the invalid area is at the lower boundary of the image to be recognized: If

or

, Then

Set to 1, that is valid; 7) Judgment method when the invalid area is at the left boundary of the image to be recognized: If

or

, Then

Set to 1, that is valid; 8) Judgment method when the invalid area is at the right boundary of the image to be recognized: If

or

, Then

Set to 1, that is valid; 9) Judgment method when the invalid area is in the central area of the image to be recognized: If

, Then

Set to 1, that is effective.

In other embodiments, the invalid region can also be re-determined according to the eight-neighborhood judgment method.

Figures 1-6 detail the fingerprint identification method of the present invention, through which the efficiency and accuracy of fingerprint identification can be improved. The functional modules of the software chip and the hardware device architecture for implementing the fingerprint identification method will be introduced below in conjunction with FIG. 7 and FIG. 8. It should be understood that the embodiments are only for illustrative purposes, and are not limited by this structure in the scope of the patent application.

FIG. 7 is a structural diagram of a fingerprint recognition chip provided by an embodiment of the present invention.

In some embodiments, the fingerprint recognition chip 200 may include a plurality of functional modules composed of code segments to realize the function of fingerprint recognition.

Referring to FIG. 7, in this embodiment, the fingerprint recognition chip 200 can be divided into a plurality of functional modules according to the functions performed by it, and the functional modules are used to perform various steps in the embodiment corresponding to FIG. 1 to Realize the function of fingerprint recognition. In this embodiment, the functional modules of the fingerprint recognition chip 200 include: an image segmentation module 201, a feature extraction module 202, an input module 203, a determination module 204, and an effective area detection module 205. The functions of each functional module will be described in detail in the following embodiments.

The image segmentation module 201 is used to obtain a fingerprint image to be identified, and segment the fingerprint image into a plurality of sub-images.

The feature extraction module 202 is used to extract the feature value of each sub-image shown.

In one embodiment, the eigenvalues are extracted using the gray gradient co-occurrence matrix as described above, and the eigenvalues include small gradient advantage, large gradient advantage, unevenness of grayscale distribution, unevenness of gradient distribution, energy , Grayscale average, gradient average, grayscale mean square error, gradient mean square error, correlation, grayscale entropy, gradient entropy, mixed entropy, difference moment, inverse difference moment.

In another embodiment, the eigenvalue is extracted using the gray level co-occurrence matrix as described above.

The input module 203 is configured to input the feature value of each sub-image into a preset machine learning model, and determine whether each sub-image is a finger image.

The determination module 204 is used to obtain the determination result of the sub-image, and when the sub-image exceeding the preset number is determined to be a finger, it is determined that the fingerprint image is a finger image.

Further, the effective area detection module 205 is configured to detect the effective area of the fingerprint in the image when it is determined that the image to be recognized is a finger image.

Specifically, the effective area detection module 205 detecting the effective area of the fingerprint includes: Filter the original gray image fingerprint to obtain the filtered gray image; Calculate the gradient of each pixel of the filtered grayscale image to obtain the gradient direction image; Divide the gradient direction image into multiple sub-blocks, and calculate the variance of each sub-block implemented; The variance of each sub-block is compared with the set threshold. If it is greater than the threshold, the sub-block is determined to be a valid area, and the flag bit of the sub-block is set to the first value; otherwise, the sub-block is determined Is an invalid area, and the flag bit of the sub-block is set to the second value. Wherein, the first value is 1, and the second value is 0.

Further, the valid area detection module 205 detecting the valid area of the fingerprint also includes after detecting the valid area of the fingerprint, for the sub-blocks determined to be invalid areas, the method further includes: according to the four-neighbor judgment method according to all the sub-blocks. The invalid area is re-determined by the values of the four adjacent area flag bits on the upper, lower, left, and right of the invalid area.

FIG. 8 is a schematic diagram of functional modules of an electronic device according to an embodiment of the present invention. The electronic device 10 includes a fingerprint collection unit 11, a memory 12, a processor 13, and a computer program 14 stored in the memory 12 and running on the processor 13, such as a fingerprint recognition program.

In this embodiment, the electronic device 10 may be, but is not limited to, a smart phone, a tablet computer, a smart industrial device, a fingerprint attendance machine, and the like.

The fingerprint collection unit 11 is used to collect fingerprint images. The fingerprint collection unit 11 can collect fingerprint images by means such as optical fingerprint collection technology, capacitive sensor fingerprint collection technology, ultrasonic fingerprint collection technology, or electromagnetic wave fingerprint collection technology.

The processor 13 implements the steps of the fingerprint identification method in the above method embodiment when the computer program 14 is executed, and is used to identify the fingerprint image collected by the fingerprint collection unit 11. Alternatively, the processor 13 executes the computer program 14 to realize the functions of the modules/units in the foregoing chip embodiment.

Exemplarily, the computer program 14 may be divided into one or more modules/units, and the one or more modules/units are stored in the memory 12 and executed by the processor 13 , To complete the present invention. The one or more modules/units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 14 in the electronic device 10. For example, the computer program 14 can be divided into modules 201-205 in FIG. 2.

Those skilled in the art can understand that the schematic diagram 3 is only an example of the electronic device 10, and does not constitute a limitation on the electronic device 10. The electronic device 10 may include more or less components than those shown in the figure, or some components may be combined. , Or different components, for example, the electronic device 10 may also include input and output devices.

The so-called processor 13 may be a central processing unit (Central Processing Unit, CPU), and may also include other general-purpose processors, digital signal processors (Digital Signal Processors, DSPs), and dedicated integrated circuits (Application Specific Integrated Circuits, ASICs). , Ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor 13 is the control center of the electronic device 10, and various interfaces and lines are used to connect various parts of the entire electronic device 10 .

The memory 12 can be used to store the computer programs 14 and/or modules/units, and the processor 13 runs or executes the computer programs and/or modules/units stored in the memory 12, and The data stored in the memory 12 is called to realize various functions of the electronic device 10. The memory body 12 may include an external storage medium or a memory body. In addition, the memory 12 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, Smart Media Card (SMC), and secure digital (Secure Digital, SD) card, flash memory card (Flash Card), at least one magnetic disk memory device, flash memory device, or other volatile solid-state memory device.

If the integrated module/unit of the electronic device 10 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 embodiments and methods, and can also be completed by instructing 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, it can implement the steps of the foregoing method embodiments. It should be noted that the content contained in the computer-readable medium can be appropriately added or deleted according to the requirements of the legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to the legislation and patent practice, the computer-readable medium Does not include electrical carrier signals and telecommunication signals.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be Modifications or equivalent replacements are made without departing from the spirit and scope of the technical solution of the present invention.

10: Electronic device 11: Fingerprint collection unit 12: Memory 13: processor 14: Computer program 200: Fingerprint recognition chip 201: Image segmentation module 202: Feature Extraction Module 203: Input Module 204: Judgment Module 205: Effective area detection module S1-S5, S201-S204, S301-S304, S401-S404: steps

FIG. 1 is a schematic flowchart of a fingerprint identification method provided by an embodiment of the present invention. FIG. 2 is a schematic flowchart of a method for extracting feature values of sub-images according to an embodiment of the present invention. Fig. 3 is a schematic diagram of a machine learning model training method provided by an embodiment of the present invention. Fig. 4 is a schematic flowchart of a method for detecting a fingerprint valid area provided by an embodiment of the present invention. Fig. 5 is a schematic diagram of the principle of a fingerprint effective area detection method provided by an embodiment of the present invention. FIG. 6 is another schematic diagram of the principle of the fingerprint effective area detection method provided by an embodiment of the present invention. Fig. 7 is a schematic diagram of a fingerprint recognition chip provided by an embodiment of the present invention. FIG. 8 is a schematic diagram of an electronic device architecture provided by an embodiment of the present invention.

S1-S5: steps

Claims (12)

A fingerprint identification method, the method includes: Acquiring the image to be recognized collected by the fingerprint acquisition device, and dividing the image into a plurality of sub-images; Extracting the feature value of each sub-image; Input the feature value of each sub-image to a preset machine learning model, and determine whether each sub-image is a finger image; Obtain the judgment result of the sub-images, and when the sub-images exceeding the preset number are judged to be finger images, it is determined that the image to be recognized is a finger image; and Perform fingerprint recognition on the finger image. According to the fingerprint identification method described in item 1 of the scope of patent application, the extracting the feature value of each sub-image includes: Generating a normalized grayscale image according to each of the sub-images; Generating a normalized gradient image of each of the sub-images; Generating a normalized gray-level gradient co-occurrence matrix according to the normalized gray-level image and the normalized gradient image; Extracting the feature value of each sub-image based on the normalized gray gradient co-occurrence matrix. For example, the fingerprint identification method described in item 2 of the scope of patent application, wherein the characteristic value includes any one or more of the following: small gradient advantage, large gradient advantage, unevenness of grayscale distribution, unevenness of gradient distribution, energy, Gray-level average, gradient average, gray-level mean square error, gradient mean square error, correlation, gray-level entropy, gradient entropy, mixed entropy, difference moment, inverse difference moment. According to the fingerprint identification method described in item 1 of the scope of patent application, the extraction of the feature value of each sub-image is based on a gray-level co-occurrence matrix. According to the fingerprint recognition method described in item 1 of the scope of patent application, the preset machine learning model includes one or more of the following: neural network model, support vector machine model, classification model based on decision tree, Bayesian Classification model. According to the fingerprint identification method described in item 5 of the scope of patent application, the training method of the machine learning model includes: Acquire a sample data set, the sample data set includes a positive sample and a negative sample, the positive sample is the feature value corresponding to the finger image, the corresponding label is the first label, and the negative sample is the feature corresponding to the non-finger image Value, the corresponding label is the second label; Dividing the sample data set into a training set and a test set; Training the machine learning model by using the training set; Use the test set to test the trained machine learning model, and adjust the parameters of the trained machine learning model according to the test results. The fingerprint identification method according to the first item of the scope of patent application, wherein, after determining that the image to be identified is a finger image, the fingerprint identification method further detects the effective area of the fingerprint in the finger image, including : Filtering the image to be recognized to obtain a filtered grayscale image; Calculating the gradient of each pixel in the filtered gray image to obtain a gradient direction image; Dividing the gradient direction image into multiple sub-blocks, and calculating the variance of each sub-block; The variance of each sub-block is compared with a set threshold. If the variance of the sub-block is greater than the threshold, the sub-block is determined to be a valid area, and the flag bit of the sub-block is set to the first A value; otherwise, the sub-block is determined to be an invalid area, and the flag bit of the sub-block is set to a second value. The fingerprint identification method described in item 7 of the scope of patent application, wherein the fingerprint identification method further includes: For the image sub-blocks determined to be invalid regions, the value of the invalid region flag bit is re-determined according to the value of the flag bit of the adjacent region of the invalid region through the four-neighbor judgment method or the eight-neighbor judgment method. A fingerprint recognition chip, the chip comprising: The image segmentation module is used to obtain the fingerprint image to be recognized, and divide the fingerprint image into a plurality of sub-images; The feature extraction module is used to extract the feature value of each sub-image shown; The input module is used for inputting the feature value of each sub-image to a preset machine learning model, and separately determining whether each sub-image is a finger image; and The judging module is used for judging the result of the sub-image. When the sub-images exceeding the preset number are judged to be a finger, the fingerprint image is determined to be a finger image. The fingerprint recognition chip described in item 9 of the scope of patent application, wherein the chip further includes: The effective area detection module is used to further detect the effective area of the fingerprint in the image after determining that the image to be recognized is a finger image, including: Filter the original gray image fingerprint to obtain the filtered gray image; Calculate the gradient of each pixel of the filtered grayscale image to obtain the gradient direction image; Divide the gradient direction image into multiple sub-blocks, and calculate the variance of each sub-block implemented; The variance of each sub-block is compared with the set threshold. If it is greater than the threshold, the sub-block is determined to be a valid area, and the flag bit of the sub-block is set to the first value; otherwise, the sub-block is determined Is an invalid area, and the flag bit of the sub-block is set to the second value. For example, the fingerprint recognition chip described in item 10 of the scope of patent application, wherein, the effective area detection module is also used to detect the effective area of the fingerprint, and use the four-neighbor judgment method for the sub-blocks determined as the invalid area Or an eight-neighborhood judgment method that re-determines the value of the invalid area flag bit according to the value of the flag bit of the adjacent area of the invalid area. An electronic device, the electronic device comprising: Fingerprint collection unit for collecting fingerprint images; processor; and A memory in which a plurality of program modules are stored, and the plurality of program modules are loaded by the processor and execute the fingerprint recognition as described in any one of items 1 to 8 of the scope of patent application method.

Images