WO2020168759A1

WO2020168759A1 - Palmprint recognition method and apparatus, computer device and storage medium

Info

Publication number: WO2020168759A1
Application number: PCT/CN2019/118424
Authority: WO
Inventors: 惠慧; 王福晴
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-02-20
Filing date: 2019-11-14
Publication date: 2020-08-27
Also published as: CN110008824A; CN110008824B

Abstract

A palmprint recognition method and apparatus, a computer device and a storage medium. The method comprises: acquiring a hand image to be recognized (S21); determining a palm area in the hand image according to a pre-trained convolutional neural network model (S22); determining a thumb root area in the hand image according to the characteristics of the palm area (S23); determining an overlapped part of the palm area and the thumb root area by performing area comparison between the palm area and the thumb root area (S24); cutting out the overlapped part from the palm area (S25); and carrying out palmprint recognition on the palm area of which the overlapped part is cut out (S26). Due to the fact that the thumb root area is cut out, interference caused by thumb deformation to palmprint recognition can be reduced, the palmprint recognition accuracy can be improved, and the palmprint matching degree is improved.

Description

Palmprint recognition method, device, computer equipment and storage medium

This application claims priority with the Chinese patent application filed on February 20, 2019 with the Chinese Patent Office, the application number is 2019101271501, and the application title is "Palmprint recognition method, device, computer equipment and storage medium", the entire content of which is incorporated by reference Incorporate in the application.

Technical field

This application relates to the field of palmprint recognition technology, and in particular to a palmprint recognition method, device, computer equipment and storage medium.

Background technique

Palm prints are the general term for all lines on the palm skin, mainly including mastoid lines, main lines and folds. The palm prints are unique, that is, the palm prints of different people are very different, and no two palms are exactly the same. Based on this feature of palmprints, identification can be performed.

At present, palmprint recognition includes contact palmprint recognition and non-contact palmprint recognition, both of which are used for identity recognition by collecting palmprint images. Among them, non-contact palmprint recognition has become a hotspot in palmprint recognition research because of its convenient and simple operation, clean and sanitary advantages. In the process of researching related technologies, the inventor found that the palmprint feature is unstable due to the deformation of the thumb during non-contact palmprint recognition, which reduces the accuracy of palmprint recognition.

Summary of the invention

The embodiments of the present application provide a palmprint recognition method, device, computer equipment, and storage medium, which can improve the accuracy of palmprint recognition.

An embodiment of the application provides a palmprint recognition method, including: acquiring a hand image to be recognized; determining a palm area in the hand image according to a pre-trained convolutional neural network model; and according to the characteristics of the palm area , Determine the thumb root area in the hand image; compare the palm area and the thumb root area to determine the overlap between the palm area and the thumb root area; The overlapping part is cut out from the palm area; palmprint recognition is performed on the palm area where the overlapping part is cut out.

An embodiment of the present application also provides a palmprint recognition device, which includes: an image acquisition module for acquiring a hand image to be recognized; a first determination module for determining a palmprint recognition device based on a pre-trained convolutional neural network model The palm area in the hand image; the second determining module is used to determine the thumb root area in the hand image according to the characteristics of the palm area; the third determining module is used to determine the palm area Perform an area comparison with the thumb root area to determine the overlap portion of the palm area and the thumb root area; a region cutting module for cutting the overlap portion from the palm area; palmprint recognition The module is used for recognizing palm prints in the palm area where the overlapping part is cut off.

An embodiment of the present application further provides a computer device, including a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor executes the aforementioned palm. Steps of pattern recognition method.

The embodiments of the present application also provide a non-volatile readable storage medium storing computer readable instructions, which when executed by one or more processors, cause the one or more processors to execute the aforementioned palm Steps of pattern recognition method.

The palmprint recognition method, device, computer equipment, and storage medium provided by the embodiments of the present application cut off the overlap between the palm area and the thumb root area before recognizing the palm area, and then perform the resection on the palm area after the overlapped portion is removed. Palmprint recognition. Since the root area of the thumb is removed, the interference caused by the deformation of the thumb on the palmprint recognition can be reduced, the accuracy of the palmprint recognition can be improved, and the palmprint matching degree can be improved.

Description of the drawings

Figure 1 is a block diagram of the internal structure of a computer device in an embodiment;

Figure 2 is a flowchart of a palmprint recognition method in an embodiment;

Figure 3 is a schematic diagram of an image collected by an image collection module in an embodiment;

Figure 4 is a schematic diagram of a hand image in an embodiment;

FIG. 5 is a schematic diagram of a process of determining the palm area in the hand image in an embodiment;

Figure 6 is a schematic diagram of a hand image in an embodiment;

Figure 7 is a schematic diagram of a hand image in an embodiment;

Figure 8 is a schematic diagram of a hand image in an embodiment;

FIG. 9 is a schematic diagram of the process of recognizing palmprints on the palm area where the overlap portion is cut in an embodiment;

Fig. 10 is a schematic structural diagram of a palmprint recognition device in an embodiment.

detailed description

In order to make the purpose, technical solutions, and advantages of this application clearer, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the application, and are not used to limit the application. It can be understood that the terms "first", "second", etc. used in this application can be used herein to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from another element.

Figure 1 is a schematic structural diagram of a computer device in an embodiment of the application. As shown in Figure 1, the computer device includes a processor, a non-volatile storage medium, a memory, and a network interface connected through a system bus. Wherein, the non-volatile storage medium of the computer device stores an operating system, a database, and computer-readable instructions. The database may store control information sequences. When the computer-readable instructions are executed by the processor, the processor can realize a A palmprint recognition method. The processor of the computer equipment is used to provide calculation and control capabilities, and supports the operation of the entire computer equipment. A computer readable instruction may be stored in the memory of the computer device. When the computer readable instruction is executed by the processor, the processor may execute a palmprint recognition method. The network interface of the computer device is used to connect and communicate with the terminal. Those skilled in the art can understand that the structure shown in FIG. 1 is only a block diagram of part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may Including more or fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.

In one embodiment, a palmprint recognition method is proposed, which can be applied to the computer device shown in FIG. 1. There are many specific application scenarios of the palmprint recognition method provided in this embodiment, and there are also many specific forms of computer equipment. For example, a company uses a non-contact palmprint recognition access control device, and employees of this company need to perform identification on the access control device when commuting. At this time, as a computer device, the access control device can use the palmprint recognition method provided in this embodiment to perform palmprint recognition, thereby realizing identity recognition. Referring to FIG. 2, the palmprint recognition method provided in this embodiment may specifically include the following steps:

S21. Obtain a hand image to be recognized; there are many triggering methods for this step, for example, when people need to perform identity recognition, trigger a button on a computer device, or perform a certain gesture operation to trigger an image acquisition module (for example, , Camera) for image acquisition. In this way, the processor in the computer device can detect the images collected by the image collection module. Among them, the image acquisition module can be a part of the computer equipment, or can be set independently of the computer equipment.

In practice, the processor in the computer device may specifically use a target detection algorithm (Single Shot MultiBox Detector, SSD for short) to detect the image collected by the image acquisition module, locate the area where the hand is, and then obtain the hand image. For example, as shown in Figure 3, the area 32 of the hand is detected in the image 31 collected by the image acquisition module, and then the area 32 of the hand is used as the hand image, which can reduce the interference of other areas on palmprint recognition. , The target detection algorithm is used to facilitate the detection of complete hand images, and incomplete hand images will be filtered out, so as to ensure that the hand image contains fingertips, finger roots and palms.

S22: Determine the palm area in the hand image according to the pre-trained convolutional neural network model; it is understandable that the palm area is the area in the hand image that mainly reflects palmprint information. It is understandable that the root area of the thumb is the area where the mark 41 in FIG. 4 is located. In practical applications, there are many ways to determine the palm area in the hand image, one of which is described below with reference to FIG. 5:

S51: Recognizing the hand image using a pre-trained convolutional neural network model to obtain the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip of any finger other than the thumb; The convolutional neural network model is trained by a training data set including a number of hand images that have marked the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip of any finger other than the thumb;

It is understandable that the position of the fingertip of any finger other than the thumb can be the position of the index finger, the middle finger, the ring finger, or the little finger. In this step, a pre-trained convolutional neural network model is used to recognize the hand image to obtain the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip of any finger other than the thumb . Taking the position of the middle finger tip as an example, the specific recognition process is explained: The Cascade convolutional neural network model is trained. The training data set includes a number of index finger root positions, pinky root positions and middle finger fingertip positions that have been marked. Hand image. After the Cascade convolutional neural network model is trained, the model is used to extract the key points of the newly input hand image, and the position of the root of the index finger, the root of the little finger, and the position of the middle finger can be obtained.

S52: Determine the palm area according to the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip. In practical applications, it can be set according to the shape of the palm area as required, for example, it can be set to any shape such as a circle, an ellipse, a square, and a rectangle.

For example, as shown in FIG. 6, the palm area is set as a square area 61 with the line AB between the base of the index finger A and the base of the little finger B as one side, and the square area 61 The center of is located on the side of the line away from the fingertip position D. Since the line AB between the base position of the index finger A and the base position B of the little finger can divide the palm area into two parts: one part is mainly the area where the four fingers other than the thumb are located, and this area is The line AB is close to the fingertip position D; the other part is mainly the area where the thumb and palm are located, and this area is the side of the line AB away from the fingertip position D. The above-mentioned square area 61 defines the approximate orientation of one side and the center, so that the only square area can be determined as the palm area.

Steps S51 and S52 provide a relatively simple method for determining the palm area. Of course, the following methods can also be used to determine the palm area: First, extract the position of the base of the index finger and the base of the little finger in the hand image according to S51 And middle finger fingertip position D. For the convenience of calculation, you can determine the tilt angle of the hand according to the line AB of the index finger base position A and the little finger base position B, the line AB and the middle finger tip position D, and rotate the hand image according to the tilt angle. Until the connection AB rotates to the horizontal direction, and the finger is up. Then, determine the center and one edge of the palm area, and then determine the palm area. Specifically: As shown in Figure 7, set the mid-perpendicular line CE of the line AB, point E is located at the midpoint of the line AB, and the length of the mid-perpendicular line CE is half the length of the line AB, and point C is taken as The midpoint of the square area, and the line AB is taken as one side of the square area to form a square area, and the square area is used as the palm area. The four vertices of the palm area are p1 (Cx-length, Cy-length), p2 (C.x+length, Cy-length), p3 (Cx-length, C.y+length) and p4 (C. x+length,C.y+length).

S23. Determine the root area of the thumb in the hand image according to the characteristics of the palm area; in practical applications, the shape of the root area of the thumb can be set as needed, for example, it can be set to a circle or an ellipse. Shape, square, rectangle and other arbitrary shapes. Taking the elliptical shape of the thumb root region as an example, a way to determine the thumb root region in the hand image will be introduced: as shown in Figure 8, the hand image is selected to meet the preset conditions The ellipse area is taken as the thumb root area; wherein the preset condition includes: the length of the long axis of the ellipse area is (1±10% of the length of the line AB )*2/5, the length of the minor axis is (1±10%)*1/4 of the length of the line AB, the center F of the elliptical area is located on the first side AG of the square area, so The first side AG is a side perpendicular to the line AB and close to the thumb. In other words, set the major axis of the elliptical area to (1±10%)*2/5 of the length of the line AB, and set the minor axis of the ellipse area to (1±10%)* of the length of the line AB 1/4, which determines the major axis and minor axis of the ellipse. The center F of the elliptical area is related to whether the hand in the hand image is the left hand or the right hand. If it is a left hand, the left side of the palm area is close to the thumb, and the center F of the ellipse area can be set on the left side of the square x-shaped palm area. If it is a right hand, the right side of the palm area is close to the thumb. Set the center F of the ellipse area on the right side of the square palm area.

Taking a hand image without rotation adjustment based on the tilt angle of the hand as an example, a method for judging whether the hand on the hand image is the left hand or the right hand is introduced: to determine whether the ordinate of the middle finger tip position is greater than the ordinate of the little finger base position. Coordinates, and determine whether the abscissa of the middle finger tip position is smaller than the abscissa of the little finger root position:

If the position of the tip of the middle finger is greater than the position of the base of the little finger, and the abscissa of the position of the middle fingertip is smaller than that of the base of the little finger, the left hand is considered;

If the position of the tip of the middle finger is greater than the position of the base of the little finger, and the abscissa of the position of the middle fingertip is greater than that of the base of the little finger, it is the right hand;

If the position of the tip of the middle finger is smaller than the base of the little finger, and the abscissa of the position of the middle finger tip is greater than the abscissa of the base of the little finger, it is the left hand;

If the position of the tip of the middle finger is smaller than the position of the base of the little finger, and the abscissa of the position of the middle finger tip is smaller than the position of the base of the little finger, it is the right hand.

It is understandable that the up and down direction of the hand can be known by judging the size relationship between the ordinate of the fingertip position of the middle finger and the ordinate of the base of the little finger. By judging the size relationship between the abscissa of the fingertip position of the middle finger and the abscissa of the base of the little finger, it can be known whether the thumb is on the left or right of the palm. Through the up and down direction of the hand and the position of the thumb relative to the palm, it can be determined whether the hand in the hand image is the left hand or the right hand. When it is clear whether the hand in the hand image is the left hand or the right hand, it can be determined which side of the square palm area the center of the ellipse area is set, and the abscissa of the center of the ellipse area can be determined.

In the above, the side of the palm area where the center of the ellipse is located is the first side, and the specific position of the center of the ellipse on the first side can be determined according to the side length of the square. For example, the ordinate of the center of the ellipse is (1±10%)*4/5 of the side length of the square. The 4/5 here is an empirical value, which can be modified according to the actual application.

S24. Determine the overlap between the palm area and the thumb root area by comparing the palm area with the thumb root area; for example, as shown in FIG. 8, the square area is the palm Area, the ellipse area is the thumb root area, and the overlap of the square area and the ellipse area is the shaded part in Figure 8. Here, the palm area and the thumb area are compared. The specific comparison method may include: determining the coordinates of each pixel in the palm area and the coordinates of each pixel in the thumb root area, and comparing the coordinates in the palm area and the thumb root area The area composed of the same pixels is the above overlapping part.

S25. Cut the overlapping part from the palm area; there are many ways to cut the overlapping part from the palm area, for example, all pixels of the overlapping part are set to 0.

S26. Perform palmprint recognition on the palm area where the overlap portion is cut off. In practical applications, there are many ways to recognize palm prints in the palm area where the overlapped part is cut. The following describes one of them with reference to FIG. 9:

S91. Determine the cosine similarity between the feature vector of the palm area where the overlap is removed and the feature vectors of multiple palmprint images stored in advance; it is understandable that the feature vector is the palm area where the overlap is removed A vector composed of multiple palmprint features of. Specifically, a machine learning model (for example, a lightweight convolutional neural network MobileNet) can be used to extract feature vectors from the palm area. It is understandable that the cosine similarity is to evaluate the similarity of the two feature vectors by calculating the cosine value of the included angle of the two feature vectors, and then to know the difference between the palm area where the overlap is removed and the pre-stored palmprint image The similarity.

S92. Determine, according to the cosine similarity, whether the palm area where the overlapping part is cut is successfully recognized. For example, if the cosine similarity between the feature vector of the palm area where the overlap is removed and the feature vector of a palmprint image stored in advance is higher than a preset threshold, it is considered that the palm area where the overlap is removed is similar to the pre-stored palmprint image. This palmprint image is sufficiently similar that the two match successfully, that is, the palm area where the overlapped part is cut is successfully recognized.

Here, the use of cosine similarity to evaluate the similarity between the palm region where the overlapped portion is cut off and the pre-stored palmprint image has the advantages of simple calculation and easy implementation. Of course, other indicators can also be used to evaluate the similarity between the two. In the palmprint recognition method provided in this embodiment, before recognizing the palm area, the overlapping part with the thumb root area is removed, and then the palmprint recognition is performed on the palm area after the overlapping part is removed. Since the root area of the thumb is removed, the interference caused by the deformation of the thumb on the palmprint recognition can be reduced, the accuracy of the palmprint recognition can be improved, and the palmprint matching degree can be improved.

As shown in FIG. 10, in one embodiment, a palmprint recognition device 100 is provided. The palmprint recognition device 100 may be integrated into the above-mentioned computer equipment, and may specifically include:

The image acquisition module 101 is used to acquire a hand image to be recognized;

The first determining module 102 is configured to determine the palm area in the hand image according to the pre-trained convolutional neural network model;

The second determining module 103 is configured to determine the thumb root area in the hand image according to the characteristics of the palm area;

The third determining module 104 is configured to determine the overlapping portion of the palm area and the thumb root area by comparing the palm area and the thumb root area;

The area cutting module 105 is used to cut the overlapped part from the palm area;

The palmprint recognition module 106 is used for recognizing palmprints on the palm area where the overlapped part is cut.

In some embodiments, the first determining module 102 includes: a first determining unit for recognizing the hand image using a pre-trained convolutional neural network model to obtain the position of the base of the index finger and the little finger The root position and the fingertip position of any finger other than the thumb; the convolutional neural network model consists of a number of marked index finger root positions, little finger root positions and any one except the thumb The training data set of the hand image of the fingertip position of the finger is obtained through training; the second determining unit is configured to determine the palm area according to the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip.

In some embodiments, the palm area is a square area with a line between the base of the index finger and the base of the little finger as one side, and the center of the square area is located on the line The side away from the fingertip position.

In some embodiments, the second determining module 103 further includes: a region selecting unit, configured to select an elliptical region that meets a preset condition in the hand image, and use the elliptical region as the thumb root region; Wherein, the preset conditions include: the length of the major axis of the elliptical area is (1±10%)*2/5 of the length of the line, and the length of the minor axis is (() of the length of the line 1±10%)*1/4, the center of the elliptical area is located on the first side of the square area, and the first side is the side perpendicular to the line and close to the thumb.

In some embodiments, the position of the center of the elliptical area on the first side is determined according to the side length of the square area.

In some embodiments, the ordinate of the center of the elliptical area is (1±10%)*4/5 of the side length of the square area.

In some embodiments, the palmprint recognition module 106 is specifically configured to: determine the cosine similarity between the feature vectors of the palm region where the overlapped portion is removed and the feature vectors of multiple prestored palmprint images; The cosine similarity determines whether the palm area where the overlap portion is cut is successfully recognized.

In the palmprint recognition device provided by the present application, before the palmprint recognition module recognizes the palm area, the area cutting module cuts off the overlap between the palm area and the thumb root area, and then the palmprint recognition module cuts off the overlapped part. Recognize palm prints in the palm area. Since the root area of the thumb is removed, the interference caused by the deformation of the thumb on the palmprint recognition can be reduced, the accuracy of the palmprint recognition can be improved, and the palmprint matching degree can be improved.

In some embodiments, a computer device is provided. The computer device includes a memory, a processor, and computer-readable instructions stored on the memory and executable on the processor, and the processor executes all The computer-readable instructions implement the following steps: obtain the hand image to be recognized; determine the palm area in the hand image according to the pre-trained convolutional neural network model; determine the palm area according to the characteristics of the palm area The thumb root area in the hand image; by comparing the palm area with the thumb root area, determine the overlap between the palm area and the thumb root area; and change the overlap from Cutting out the palm area; and performing palm print recognition on the palm area where the overlapping part is cut out. In some embodiments, the determining the palm area in the hand image according to a pre-trained convolutional neural network model executed by the processor includes: using a pre-trained convolutional neural network model to Recognizing the hand image, the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip of any finger other than the thumb are obtained; the convolutional neural network model is marked by including several The index finger root position, the little finger root position, and the fingertip position of any finger other than the thumb are trained on the training data set of hand images; according to the index finger root position, the little finger root position and the position The fingertip position determines the palm area. In some embodiments, the palm area is a square area with a line between the base of the index finger and the base of the little finger as one side, and the center of the square area is located on the line The side away from the fingertip position.

In some embodiments, the step of determining the root region of the thumb in the hand image performed by the processor includes: selecting an elliptical region that meets a preset condition in the hand image, and combining the ellipse The area is taken as the root area of the thumb; wherein the preset conditions include: the length of the long axis of the elliptical area is (1±10%)*2/5 of the length of the line, and the length of the short axis Is (1±10%)*1/4 of the length of the line, the center of the elliptical area is located on the first side of the square area, and the first side is perpendicular to the line and close to The side of the thumb. In some embodiments, the position of the center of the elliptical area on the first side is determined according to the side length of the square area. In some embodiments, the ordinate of the center of the elliptical area is (1±10%)*4/5 of the side length of the square area. In some embodiments, the step of recognizing palm prints on the palm area from which the overlapped portion is cut off by the processor includes: determining that the feature vector of the palm area from which the overlap portion is cut off is different from a plurality of pre-stored palms. The cosine similarity between the feature vectors of the pattern image; according to the cosine similarity, it is determined whether the palm area where the overlapped part is cut off is successfully recognized.

The beneficial effects of the computer equipment provided in this application are the same as the palmprint recognition method and device described above, and will not be repeated here.

In one embodiment, a non-volatile readable storage medium storing computer readable instructions is provided. When the computer readable instructions are executed by one or more processors, the one or more processors execute the following Steps: Obtain the hand image to be recognized; determine the palm area in the hand image according to the pre-trained convolutional neural network model; determine the thumb root in the hand image according to the characteristics of the palm area Area; by comparing the palm area and the thumb root area to determine the overlap of the palm area and the thumb root area; cut the overlap from the palm area; The palm area of the overlapping part is cut out for palmprint recognition. In some embodiments, the determination of the palm area in the hand image based on the pre-trained convolutional neural network model executed by one or more processors includes: using the pre-trained convolutional neural network model to perform The hand image is identified to obtain the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip of any finger other than the thumb; the convolutional neural network model consists of a number of labeled The training data set of hand images of the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip of any finger other than the thumb is obtained by training; according to the position of the base of the index finger, the position of the base of the little finger, and The fingertip position determines the palm area. In some embodiments, the palm area is a square area with a line between the base of the index finger and the base of the little finger as one side, and the center of the square area is located on the line The side away from the fingertip position. In some embodiments, the step performed by one or more processors to determine the thumb root region in the hand image includes: selecting an elliptical region that meets a preset condition in the hand image, and combining the An elliptical area is used as the root area of the thumb; wherein the preset conditions include: the length of the long axis of the elliptical area is (1±10%)*2/5 of the length of the connecting line, and the short axis is The length is (1±10%)*1/4 of the length of the line, the center of the elliptical area is located on the first side of the square area, and the first side is perpendicular to the line and Near the side of the thumb. In some embodiments, the position of the center of the elliptical area on the first side is determined according to the side length of the square area. In some embodiments, the ordinate of the center of the elliptical area is (1±10%)*4/5 of the side length of the square area. In some embodiments, the steps performed by one or more processors to perform palmprint recognition on the palm region from which the overlapping part is cut off include: determining that the feature vector of the palm region from which the overlapping part is cut off is different from a plurality of pre-stored The cosine similarity between the feature vectors of the palmprint images; according to the cosine similarity, it is determined whether the palm area where the overlapped part is cut is successfully recognized.

The beneficial effects of the non-volatile readable storage medium provided in the present application are the same as those of the palmprint recognition method and device described above, and will not be repeated here. A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing relevant hardware through a computer program. The computer program can be stored in a computer readable storage medium. When executed, it may include the processes of the above-mentioned method embodiments. Among them, the aforementioned storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disc, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM), etc.

The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, All should be considered as the scope of this specification. The above-mentioned embodiments only express several implementation manners of this application, and their descriptions are more specific and detailed, but they should not be construed as limiting the scope of this application. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can be made, and these all fall within the protection scope of this application. Therefore, the scope of protection of the patent of this application shall be subject to the appended claims.

Claims

A palmprint recognition method, the method includes:

Obtain the hand image to be recognized;

Determining the palm area in the hand image according to the pre-trained convolutional neural network model;

Determining the thumb root area in the hand image according to the characteristics of the palm area;

By performing an area comparison between the palm area and the root area of the thumb to determine the overlap portion of the palm area and the root area of the thumb;

Cutting the overlapping part from the palm area;

The palmprint recognition is performed on the palm area where the overlapping part is cut off.
The method according to claim 1, wherein the determining the palm area in the hand image according to a pre-trained convolutional neural network model comprises: performing a pre-trained convolutional neural network model on the hand image Recognition, the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip of any finger other than the thumb are obtained; the convolutional neural network model includes a number of index finger root positions marked , The base of the little finger and the fingertip position of any finger other than the thumb are trained on the training data set of hand images; according to the base position of the index finger, the base position of the little finger and the position of the fingertip , Determine the palm area.
The method according to claim 2, wherein the palm area is a square area with a line between the base of the index finger and the base of the little finger as one side, and the center of the square area is located at the The connecting line is far away from the fingertip position.
The method according to claim 3, the determining the thumb root region in the hand image comprises:

In the hand image, an elliptical area that meets a preset condition is selected, and the elliptical area is used as the thumb root area; wherein, the preset condition includes: the length of the long axis of the elliptical area is The length of the line is (1±10%)*2/5, the length of the minor axis is (1±10%)*1/4 of the length of the line, and the center of the ellipse is located in the square On the first side of the area, the first side is the side perpendicular to the line and close to the thumb.
The method according to claim 4, wherein the position of the center of the elliptical area on the first side is determined according to the side length of the square area.
The method according to claim 4, wherein the ordinate of the center of the elliptical area is (1±10%)*4/5 of the side length of the square area.
The method according to claim 1, performing palmprint recognition on the palm region where the overlapping part is cut out, comprising: determining the feature vector of the palm region where the overlapping part is cut out and the feature vector of a plurality of pre-stored palmprint images respectively The cosine similarity between the two; according to the cosine similarity, it is determined whether the palm area where the overlapping part is cut is successfully recognized.
A palmprint recognition device, the device comprising:

The image acquisition module is used to acquire the hand image to be recognized;

The first determining module is configured to determine the palm area in the hand image according to the pre-trained convolutional neural network model;

The second determining module is configured to determine the thumb root area in the hand image according to the characteristics of the palm area;

A third determining module, configured to determine the overlapping part of the palm area and the thumb root area by comparing the palm area and the thumb root area;

A region cutting module for cutting the overlapping part from the palm region;

The palmprint recognition module is used for recognizing palmprints in the palm area where the overlapping part is cut.
The apparatus according to claim 8, wherein the first determining module comprises:

The first determining unit is used to recognize the hand image using a pre-trained convolutional neural network model to obtain the position of the base of the index finger, the position of the base of the little finger, and any finger other than the thumb The position of the fingertip; the convolutional neural network model consists of a training data set of hand images that have been marked with the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip of any finger except the thumb Trained

The second determining unit is configured to determine the palm area according to the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip.
The device according to claim 9, wherein the palm area is a square area with a line between the base of the index finger and the base of the little finger as a side, and the center of the square area is located at the The connecting line is far away from the fingertip position.
The device according to claim 10, the second determining module further comprises:

An area selection unit, configured to select an elliptical area that meets a preset condition from the hand image, and use the elliptical area as the thumb root area; wherein the preset condition includes: The length of the major axis is (1±10%)*2/5 of the length of the line, and the length of the minor axis is (1±10%)*1/4 of the length of the line, the elliptical area The center of is located on the first side of the square area, and the first side is the side perpendicular to the line and close to the thumb.
The device according to claim 11, wherein the position of the center of the elliptical area on the first side is determined according to the side length of the square area.
The device according to claim 11, the ordinate of the center of the elliptical area is (1±10%)*4/5 of the side length of the square area.
8. The device according to claim 8, wherein the palmprint recognition module is specifically configured to determine the cosine similarity between the feature vectors of the palm region where the overlapped portion is cut off and the feature vectors of multiple prestored palmprint images. ; According to the cosine similarity, it is determined whether the recognition of the palm area where the overlapping part is cut off is successful.
A computer device includes a memory and a processor, and the memory stores computer readable instructions, and when the computer readable instructions are executed by the processor, the processor executes the steps of the palmprint recognition method, including : Obtain the hand image to be recognized; determine the palm area in the hand image according to the pre-trained convolutional neural network model; determine the thumb root area in the hand image according to the characteristics of the palm area By comparing the area of the palm area and the root area of the thumb to determine the overlap portion of the palm area and the root area of the thumb; cutting the overlap portion from the palm area; Palm print recognition is performed on the palm area of the overlapping part.
The computer device according to claim 15, when the computer-readable instructions are executed by the processor, the processor executes the pre-trained convolutional neural network model to determine the content in the hand image The palm area includes: recognizing the hand image using a pre-trained convolutional neural network model to obtain the position of the base of the index finger, the position of the base of the little finger, and the finger of any finger except the thumb Tip position; the convolutional neural network model is trained by a training dataset that includes a number of hand images that have marked the position of the base of the index finger, the position of the base of the little finger, and the position of the tip of any finger except the thumb ; Determine the palm area according to the position of the base of the index finger, the position of the base of the little finger and the position of the fingertip.
The computer device according to claim 16, wherein when the computer-readable instructions are executed by the processor to implement the method, the palm area is one between the position of the base of the index finger and the base of the little finger. The connecting line is a square area with one side, and the center of the square area is located on the side of the connecting line away from the fingertip position.
A non-volatile readable storage medium storing computer readable instructions. When the computer readable instructions are executed by one or more processors, the one or more processors execute the steps of the palmprint recognition method, including : Obtain the hand image to be recognized; determine the palm area in the hand image according to the pre-trained convolutional neural network model; determine the thumb root area in the hand image according to the characteristics of the palm area By comparing the area of the palm area and the root area of the thumb to determine the overlap portion of the palm area and the root area of the thumb; cutting the overlap portion from the palm area; Palm print recognition is performed on the palm area of the overlapping part.
The storage medium of claim 18, when the computer-readable instructions are executed by one or more processors, the one or more processors execute the pre-trained convolutional neural network model to determine the hand The palm area in the image includes: recognizing the hand image using a pre-trained convolutional neural network model to obtain the position of the base of the index finger, the position of the base of the little finger, and any other than the thumb The position of the fingertip of a finger; the convolutional neural network model is trained by including several hand images that have marked the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip of any finger except the thumb The data set is obtained by training; the palm area is determined according to the position of the base of the index finger, the position of the base of the little finger, and the position of the fingertip.
The storage medium according to claim 19, when the computer-readable instructions are executed by the processor to implement the method, the palm area is one between the position of the base of the index finger and the base of the little finger. The connecting line is a square area with one side, and the center of the square area is located on the side of the connecting line away from the fingertip position.