Attendance card punching system based on image recognition

Abstract

The invention discloses an attendance checking and card punching system based on image recognition, relates to the technical field of intelligent attendance checking, and solves the technical problem of comprehensive calculation and analysis of a proportion value of facial key feature points extracted by using an image technology, and the attendance checking and card punching system comprises the following steps: preprocessing collected facial features and extracting key points; a comprehensive proportion calculation value is obtained through the eye-nose, nose-mouth and eye-mouth pixel point proportion values for analysis; storing a comprehensive proportion storage calculation numerical value obtained by the pixel point proportion value numerical value of the registered attendance personnel in a database; obtaining a comprehensive feature adaptation degree value through the comprehensive proportion calculation value, the comprehensive proportion storage calculation value and the comprehensive feature proportion coefficient, comparing the comprehensive feature adaptation degree value with a preset threshold value, and judging whether matching succeeds or not; and dividing corresponding security threat signals according to the number of continuous failures and carrying out early warning.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent attendance, and in particular relates to an attendance punching system based on image recognition. Background Art

[0002] With the continuous development of artificial intelligence technology, image recognition, as an important part of artificial intelligence, has become an efficient means of identity authentication in attendance, which can significantly improve the security and convenience of attendance punching systems. According to the facial feature key points, eyes, nose, nose and mouth, and eye and mouth pixel point ratio values, the corresponding weights are assigned to obtain the comprehensive ratio calculation value, which is combined with the comprehensive ratio storage calculation value and the comprehensive ratio coefficient value for analysis and matching. The security threat signal is activated by the number of recognition failures, which greatly simplifies the traditional card or password input method and improves the accuracy and security of the system.

[0003] Although the existing attendance punching system has achieved attendance punching to a certain extent, it usually relies on physical media or password input, which are at risk of being copied or shared. The lack of image recognition to extract key points of facial features makes it difficult to combine various comprehensive features to calculate numerical values ​​for analysis and matching, and sends different degrees of security threat signals for the number of consecutive recognition failures. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an attendance punching system based on image recognition, which is used to solve the following technical problems:

[0005] Although the existing attendance punching system has achieved attendance punching to a certain extent, it usually relies on physical media or password input, which are at risk of being copied or shared. The lack of image recognition to extract key points of facial features makes it difficult to combine various comprehensive features to calculate numerical values ​​for analysis and matching, and sends different degrees of security threat signals for the number of consecutive recognition failures.

[0006] To solve the above problems, the first aspect of the present invention provides an attendance punching system based on image recognition, comprising the following modules:

[0007] Image acquisition module: uses a high-resolution camera to extract facial images of attendance personnel and attendance timestamp data, and transmits the data to the image recognition module;

[0008] Image recognition module: perform image preprocessing on the extracted facial images of attendance personnel and attendance timestamp data and extract key feature points of facial images, obtain comprehensive ratio calculation values ​​through the pixel ratio values ​​of eyes and nose, nose and mouth, and eyes and mouth to analyze and identify the extracted data;

[0009] Image storage module: stores the facial images of the registered attendance personnel, obtains the comprehensive ratio storage calculation value set {A1, A2, A3, ..., An} according to the pixel ratio values ​​of the eyes, nose, nose and mouth, and eyes and mouth of the registered attendance personnel, and stores the set in the database;

[0010] Image matching module: Based on the comprehensive feature ratio coefficient obtained by the comprehensive weight difference between the extracted facial feature key points and the stored facial feature key points, the comprehensive feature fitness value is obtained through the comprehensive ratio calculation value, the comprehensive ratio storage calculation value and the comprehensive feature ratio coefficient, and compared with the preset threshold to determine whether the match is successful;

[0011] Image security module: Set the threshold for the number of consecutive recognition failures, classify the security threat level according to the number of consecutive recognition failures, and issue a security warning.

[0012] As a further solution of the present invention: the image acquisition extracts the facial image and attendance timestamp data of the attendance personnel through a high-resolution camera, comprising the following steps:

[0013] The facial image data of the attendance personnel is obtained through a high-resolution camera, and the system automatically records the current timestamp as the attendance time, and transmits the facial image and attendance time data of the attendance personnel to the image recognition module.

[0014] As a further solution of the present invention: the image recognition module performs image preprocessing on the collected facial image and attendance time data of the attendance personnel, including image grayscale and histogram equalization, including the following steps:

[0015] S201: receiving the facial image and attendance time data of the attendance personnel from the image acquisition module, and converting the color image into a grayscale image through the cvtColor function in the OpenCV cross-platform computer vision and machine learning software library;

[0016] S202: Perform histogram equalization on the grayscale image using the equalHist function in the OpenCV cross-platform computer vision and machine learning software library.

[0017] As a further solution of the present invention: the extraction and detection of key feature points of facial images comprises the following steps:

[0018] By combining OpenCV and D lib library, the key facial feature points of the attendance personnel, including eyes, nose and mouth, are detected. According to the extracted facial nose key point coordinates A (x1nose, y1nose, z1nose) and B (x2nose, y2nose, y3nose), facial eye key point coordinates C (x1eye, y1eye, z1eye) and D (x2eye, y2eye, z2eye), facial mouth key point coordinates E (x1mouth, y1mouth, z1mouth) and F (x2mouth, y2mouth, z2mouth), origin coordinates, the pixel ratio values ​​of the eyes and nose, nose and mouth, and eyes and mouth are calculated, and different weights are assigned to the ratio values ​​to construct a comprehensive ratio calculation value;

[0019] According to the ratio of the distance between the facial nose key point coordinates A (x1nose, y1nose, z1nose) and B (x2nose, y2nose, y3nose) and the average length of the distance between the two points AB and the origin coordinates, combined with the distance between the facial eye key point coordinates C (x1eye, y1eye, z1eye) and D (x2eye, y2eye, z2eye) and the average length of the distance between the two points CD and the origin coordinates, the eye-nose pixel ratio value calculation formula is obtained by comparing the two:

[0020]

[0021] Among them, O nose L is the average length of the distance between points A and B and the origin of the coordinate system. nose is the ratio of the distance between the two key points AB of the nose to the average length, O eye L is the average length of the distance between points C and D and the origin of the coordinate system. eye is the ratio of the distance between the eye key point CD and the average value, R, G and B are the red, green and blue points of the color image, r1 is the calculated value of the ratio of eye and nose pixels;

[0022] According to the ratio of the distance between the facial nose key point coordinates A (x1nose, y1nose, z1nose) and B (x2nose, y2nose, y3nose) and the average length of the distance between the two points AB and the origin coordinates, combined with the distance between the facial mouth key point coordinates E (x1mouth, y1mouth, z1mouth) and F (x2mouth, y2mouth, z2mouth) and the average length of the distance between the two points EF and the origin coordinates, the nose and mouth pixel ratio value calculation formula is obtained by comparing the two:

[0023]

[0024] Among them, O mouth L is the average length of the distance between points E and F and the origin of the coordinate system. mouth is the ratio of the distance between the two key points EF of the mouth to the average length, and r2 is the calculated value of the ratio of the pixels of the nose and mouth;

[0025] According to the ratio of the distance between the facial eye key point coordinates C (x1eye, y1eye, z1eye) and D (x2eye, y2eye, z2eye) and the average length of the distance between CD and the origin coordinates, combined with the distance between the facial mouth key point coordinates E (x1mouth, y1mouth, z1mouth) and F (x2mouth, y2mouth, z2mouth) and the average length of the distance between EF and the origin coordinates, the eye-mouth pixel ratio value calculation formula is obtained by comparing the two:

[0026]

[0027] Among them, L eye L is the distance between the eye key points CD. mouth is the distance between the mouth key points EF, and r3 is the calculated value of the eye-mouth pixel ratio;

[0028] The comprehensive ratio calculation formula is obtained by assigning different weights to the pixel ratio values ​​of eyes and nose, nose and mouth, and eyes and mouth:

[0029] a=(r1×W1+r2×W2+r3×W3)×100%

[0030] Among them, a is the calculated value of the comprehensive ratio, r1 is the calculated value of the ratio of eye and nose pixels, r2 is the calculated value of the ratio of nose and mouth pixels, r3 is the calculated value of the ratio of eye and mouth pixels, W1, W2, and W3 are the corresponding weights, W1 is 30%, W2 is 20%, and W3 is 50%.

[0031] As a further solution of the present invention: the image storage module stores the facial features of the registered attendance personnel, including the following steps:

[0032] According to the key point ratio values ​​of the eyes and nose, nose and mouth, and eyes and mouth of the registered attendance personnel, a comprehensive ratio storage calculation value set {A1, A2, A3, ..., An} is obtained and stored in the database;

[0033] The comprehensive ratio storage calculation formula is obtained through the stored key point ratio values ​​of eyes and nose, nose and mouth, and eyes and mouth:

[0034] A i =(R1×W1+R2×W2+R3×W3)×100%

[0035] Among them, each A i Represents the calculated value of the comprehensive ratio storage of the i-th registered attendance personnel, R1 is the calculated value of the eye-nose pixel ratio, R2 is the calculated value of the nose-mouth pixel ratio, R3 is the calculated value of the eye-mouth pixel ratio, i∈(1, 2, ..., n), n is the total number of attendance personnel, W1, W2, W3 are the corresponding weights, W1 is 30%, W2 is 20%, and W3 is 50%.

[0036] As a further solution of the present invention: the image matching module obtains the comprehensive feature ratio coefficient according to the comprehensive weight difference between the extracted facial feature key points and the stored facial feature key points, comprising the following steps:

[0037] The calculation formula for comprehensive feature fitness is obtained based on the comprehensive ratio calculation value and the comprehensive ratio storage calculation value:

[0038]

[0039] Among them, α is the comprehensive characteristic proportional coefficient, r recognized is the extracted facial feature key point vector, R stored is the stored facial feature key point vector, R1, R2, R3 are the stored calculated values ​​of eye-nose, nose-mouth and eye-mouth pixel ratios, r1, r2, r3 are the calculated values ​​of eye-nose, nose-mouth and eye-mouth pixel ratios, W1, W2, W3 are the corresponding weights, W1 is 30%, W2 is 20%, and W3 is 50%.

[0040] As a further solution of the present invention: obtaining a comprehensive feature fitness value through comprehensive ratio calculation value, comprehensive ratio storage calculation value and comprehensive feature ratio coefficient, and comparing it with a preset threshold value to determine whether the match is successful, including the following steps:

[0041]

[0042] Among them, α is the comprehensive feature ratio coefficient, a is the comprehensive ratio calculation value, Ai is the comprehensive ratio storage calculation value of the registered attendance personnel, and P is the comprehensive feature fitness calculation value;

[0043] If P ≥ T, the match is successful and the attendance punch-in is recorded successfully;

[0044] If P < T, the match fails, and a security threat warning is issued according to the set number of consecutive failures;

[0045] Among them, P is the calculated value of comprehensive feature fitness, and T is the preset threshold.

[0046] As a further solution of the present invention: the image security module divides the security threat level according to the number of consecutive recognition failures, including the following steps:

[0047] Set a threshold for the number of consecutive recognition failures, classify the security threat level according to the number of consecutive recognition failures, and issue a security warning;

[0048] According to the set threshold of consecutive identification failures, the security threat level can be divided into three levels:

[0049] If the number of consecutive recognition failures is 1-2 times, a low threat signal is issued;

[0050] If the number of consecutive identification failures is 3-4, a medium threat signal is issued;

[0051] If the number of consecutive identification failures is 5 or more, a high threat signal will be issued.

[0052] Beneficial effects of the present invention:

[0053] The invention discloses an attendance punching system based on image recognition, relates to the field of intelligent attendance technology, solves the technical problem of comprehensive calculation and analysis of the ratio value of key facial feature points extracted by using image technology, obtains the comprehensive ratio calculation value through the pixel ratio value of the eyes, nose, nose and mouth, and analyzes the extracted data, and realizes the simplification of the cumbersome process of traditional manual attendance; the comprehensive ratio storage calculation value obtained by the pixel ratio value of the registered attendance personnel is stored in a database; the comprehensive feature fitness value is obtained according to the comprehensive ratio calculation value, the comprehensive ratio storage calculation value and the comprehensive feature ratio coefficient, and the comprehensive feature fitness value is compared with the preset value to judge whether the match is successful. If the match is successful, the attendance punching is successful; if the match fails, different degrees of security threat warning signals are automatically issued according to the number of consecutive failures. Automated attendance punching is realized, and the efficiency and accuracy of attendance management are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 Schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] See also Figure 1 As shown, the present invention is an attendance punching system based on image recognition, comprising the following modules:

[0058] Image acquisition module: uses a high-resolution camera to extract the facial features and attendance timestamp data of the attendance personnel, and transmits the data to the image recognition module;

[0059] Image recognition module: perform image preprocessing on the extracted facial features and attendance timestamp data of the attendance personnel and extract key feature points of the facial image to detect the extracted data. The comprehensive ratio calculation value is obtained through the pixel ratio values ​​of the eyes and nose, nose and mouth, and eyes and mouth to analyze and identify the extracted data.

[0060] Image storage module: stores the facial features of the registered attendance personnel, obtains the comprehensive ratio storage calculation value set {A1, A2, A3, ..., An} according to the pixel ratio values ​​of the registered attendance personnel's eyes and nose, nose and mouth, and eyes and mouth, and stores the set in the database;

[0061] Image matching module: Based on the comprehensive feature ratio coefficient obtained by the comprehensive weight difference between the extracted facial feature key points and the stored facial feature key points, the comprehensive feature fitness value is obtained through the comprehensive ratio calculation value, the comprehensive ratio storage calculation value and the comprehensive feature ratio coefficient, and compared with the preset threshold to determine whether the match is successful;

[0062] Image security module: Set the threshold for the number of consecutive recognition failures, classify the security threat level according to the number of consecutive recognition failures, and issue a security warning.

[0063] Specifically, the collected facial features are preprocessed and key points are extracted, and the comprehensive ratio calculation value is obtained through the pixel ratio values ​​of the eyes, nose, nose and mouth, and the eyes and mouth to analyze and identify the extracted data; the comprehensive ratio storage calculation value obtained by the pixel ratio values ​​of the eyes, nose, nose and mouth of the registered attendance personnel is stored in the database; the comprehensive feature fitness value is obtained according to the comprehensive ratio calculation value, the comprehensive ratio storage calculation value and the comprehensive feature ratio coefficient value, and the comprehensive feature fitness value is compared with the preset value to determine whether the match is successful. If the comprehensive feature fitness value is equal to or greater than the preset value, the attendance punching is successful; if the comprehensive feature fitness value is less than the preset value, different degrees of security threat warning signals are issued according to the number of consecutive failures. The security and convenience of the attendance punching system are improved, and the accuracy of the attendance data is also improved.

[0064] In one embodiment of the present invention, the image acquisition extracts the facial image and attendance timestamp data of the attendance personnel through a high-resolution camera, including the following steps:

[0065] The facial image of the attendance personnel is obtained through a high-resolution camera. The system records the current timestamp as the attendance time in real time and transmits the facial image and attendance time data of the attendance personnel to the image recognition module.

[0066] Specifically, a high-resolution camera is used to ensure that the facial image of the attendance personnel is clearly visible, which helps to improve the accuracy of subsequent image processing and recognition. The system uses the camera to capture the facial image of the attendance personnel in the field of view in real time to ensure that no attendance record is missed. The system also automatically records the current timestamp as the attendance time, so that each clock-in has an accurate time record, and transmits the facial image to the image recognition module for subsequent processing.

[0067] In one embodiment of the present invention, the image recognition module performs image preprocessing on the collected facial image and attendance time data of the attendance personnel, including image grayscale and histogram equalization, including the following steps:

[0068] S201: receiving the facial image and attendance time data of the attendance personnel from the image acquisition module, and converting the color image into a grayscale image through the cvtColor function in the OpenCV cross-platform computer vision and machine learning software library;

[0069] S202: Perform histogram equalization on the grayscale image using the equalHist function in the OpenCV cross-platform computer vision and machine learning software library.

[0070] Specifically, the received facial images and attendance time data are converted to grayscale and histogram equalized through various functions in the OpenCV cross-platform computer vision and machine learning software library. The cvtColor function is used to convert the color image into a gray image, which helps to reduce the computational complexity and eliminate the interference of color on the recognition algorithm, and converts the color value of each pixel into the corresponding grayscale value. The equalizeHist function is used to improve the contrast of the image so that the grayscale distribution in different brightness areas is more uniform, which is especially important for images captured in uneven lighting environments.

[0071] In one embodiment of the present invention, the step of extracting and detecting key feature points of a facial image comprises the following steps:

[0072] By combining OpenCV and D lib library, the key facial feature points of the attendance personnel, including eyes, nose and mouth, are detected. According to the extracted facial nose key point coordinates A (x1nose, y1nose, z1nose) and B (x2nose, y2nose, y3nose), facial eye key point coordinates C (x1eye, y1eye, z1eye) and D (x2eye, y2eye, z2eye), facial mouth key point coordinates E (x1mouth, y1mouth, z1mouth) and F (x2mouth, y2mouth, z2mouth), origin coordinates, the pixel ratio values ​​of the eyes and nose, nose and mouth, and eyes and mouth are calculated, and different weights are assigned to the ratio values ​​to construct a comprehensive ratio calculation value;

[0073] According to the ratio of the distance between the facial nose key point coordinates A (x1nose, y1nose, z1nose) and B (x2nose, y2nose, y3nose) and the average length of the distance between the two points AB and the origin coordinates, combined with the distance between the facial eye key point coordinates C (x1eye, y1eye, z1eye) and D (x2eye, y2eye, z2eye) and the average length of the distance between the two points CD and the origin coordinates, the eye-nose pixel ratio value calculation formula is obtained by comparing the two:

[0074]

[0075] Among them, O nose L is the average length of the distance between points A and B and the origin of the coordinate system. nose is the ratio of the distance between the two key points AB of the nose to the average length, O eye L is the average length of the distance between points C and D and the origin of the coordinate system. eyeis the ratio of the distance between the eye key point CD and the average value, R, G and B are the red, green and blue points of the color image, r1 is the calculated value of the ratio of eye and nose pixels;

[0076] According to the ratio of the distance between the facial nose key point coordinates A (x1nose, y1nose, z1nose) and B (x2nose, y2nose, y3nose) and the average length of the distance between the two points AB and the origin coordinates, combined with the distance between the facial mouth key point coordinates E (x1mouth, y1mouth, z1mouth) and F (x2mouth, y2mouth, z2mouth) and the average length of the distance between the two points EF and the origin coordinates, the nose and mouth pixel ratio value calculation formula is obtained by comparing the two:

[0077]

[0078] Among them, O mouth L is the average length of the distance between points E and F and the origin of the coordinate system. mouth is the ratio of the distance between the two key points EF of the mouth to the average length, and r2 is the calculated value of the ratio of the pixels of the nose and mouth;

[0079] According to the ratio of the distance between the facial eye key point coordinates C (x1eye, y1eye, z1eye) and D (x2eye, y2eye, z2eye) and the average length of the distance between CD and the origin coordinates, combined with the distance between the facial mouth key point coordinates E (x1mouth, y1mouth, z1mouth) and F (x2mouth, y2mouth, z2mouth) and the average length of the distance between EF and the origin coordinates, the eye-mouth pixel ratio value calculation formula is obtained by comparing the two:

[0080]

[0081] Among them, L eye L is the distance between the eye key points CD. mouth is the distance between the mouth key points EF, and r3 is the calculated value of the eye-mouth pixel ratio;

[0082] The comprehensive ratio calculation formula is obtained by assigning different weights to the pixel ratio values ​​of eyes and nose, nose and mouth, and eyes and mouth:

[0083] a=(r1×W1+r2×W2+r3×W3)×100%

[0084] Among them, a is the calculated value of the comprehensive ratio, r1 is the calculated value of the ratio of eye and nose pixels, r2 is the calculated value of the ratio of nose and mouth pixels, r3 is the calculated value of the ratio of eye and mouth pixels, W1, W2, and W3 are the corresponding weights, W1 is 30%, W2 is 20%, and W3 is 50%.

[0085] Specifically, the Haar cascade classifier in the OpenCV library is used for face detection to locate the face area in the image. These classifiers have been trained and are ready to recognize the approximate position of the face. In the detected face area, the D lib library is used to further accurately locate the key feature points of the face. D lib provides a pre-trained facial labeling model to extract the coordinates of the key feature points of the eyes, nose and mouth, and the origin coordinates. According to the extracted coordinates, the pixel ratio values ​​of the eyes and nose, nose and mouth, and eyes and mouth can be calculated. By assigning corresponding weights to the comprehensive ratio calculation value, it is very important to obtain these ratio information for face recognition.

[0086] In one embodiment of the present invention, the image storage module stores the facial features of the registered attendance personnel, including the following steps:

[0087] The facial features of the registered attendance personnel are stored, and a comprehensive ratio storage calculation value set {A1, A2, A3, ..., An} is obtained according to the pixel ratio values ​​of the eyes, nose, nose and mouth, and eyes and mouth of the registered attendance personnel;

[0088] A i =(R1×W1+R2×W2+R3×W3)×100%

[0089] Among them, each A i Represents the calculated value of the comprehensive ratio storage of the i-th registered attendance personnel, R1 is the calculated value of the eye-nose pixel ratio, R2 is the calculated value of the nose-mouth pixel ratio, R3 is the calculated value of the eye-mouth pixel ratio, i∈(1, 2, ..., n), n is the total number of attendance personnel, W1, W2, W3 are the corresponding weights, W1 is 30%, W2 is 20%, and W3 is 50%.

[0090] Specifically, the attendance personnel need to provide a facial image when registering, and the comprehensive proportion storage calculation value is obtained according to the pixel ratio calculation values ​​of the eyes and nose, nose and mouth, and eyes and mouth entered by the attendance personnel, and a comprehensive proportion storage calculation value set {A1, A2, A3, ..., An} is generated and matched with the comprehensive proportion calculation value collected in the image recognition module. These proportion values ​​are the key to achieving accurate face recognition and are crucial to improving the accuracy and reliability of the attendance punching system.

[0091] In one embodiment of the present invention, the image matching module obtains the comprehensive feature ratio coefficient according to the comprehensive weight difference between the extracted facial feature key points and the stored facial feature key points, comprising the following steps:

[0092] The calculation formula for comprehensive feature fitness is obtained based on the comprehensive ratio calculation value and the comprehensive ratio storage calculation value:

[0093]

[0094] Among them, α is the comprehensive characteristic proportional coefficient, r recognized is the extracted facial feature key point vector, R stored is the stored facial feature key point vector, R1, R2, R3 are the stored calculated values ​​of eye-nose, nose-mouth and eye-mouth pixel ratios, r1, r2, r3 are the calculated values ​​of eye-nose, nose-mouth and eye-mouth pixel ratios, W1, W2, W3 are the corresponding weights, W1 is 30%, W2 is 20%, and W3 is 50%.

[0095] Specifically, the comprehensive feature ratio coefficient value is obtained by the comprehensive weight difference between the extracted facial feature key points and the stored facial feature key points of eyes, nose, nose and mouth, and eyes and mouth pixels, thereby improving the accuracy of image processing and facial recognition.

[0096] In one embodiment of the present invention, a comprehensive feature fitness value is obtained by calculating a comprehensive ratio value, storing a comprehensive ratio value and a comprehensive feature ratio coefficient, and compared with a preset threshold value to determine whether the match is successful, including the following steps:

[0097]

[0098] Among them, α is the comprehensive feature ratio coefficient, a is the comprehensive ratio calculation value, Ai is the comprehensive ratio storage calculation value of the registered attendance personnel, and P is the comprehensive feature fitness calculation value;

[0099] If P ≥ T, the match is successful and the attendance punch-in is recorded successfully;

[0100] If P < T, the match fails, and a security threat warning is issued according to the set number of consecutive failures;

[0101] Among them, P is the calculated value of the comprehensive feature fitness, and T is the preset threshold, which is a constant of 0.7 in the calculation formula.

[0102] Specifically, the comprehensive feature proportion coefficient value, the comprehensive proportion calculation value and the comprehensive proportion storage calculation value are used to obtain the comprehensive feature fitness calculation value, which is compared with the preset threshold value. If the match is successful, the attendance status of the attendance personnel is recorded and the attendance punch-in is marked as successful; if the match fails, a safety warning is issued according to the set number of consecutive failures, thereby ensuring the accuracy and security of the attendance punch-in system.

[0103] In one embodiment of the present invention, the image security module divides the security threat level according to the number of consecutive recognition failures, including the following steps:

[0104] Set a threshold for the number of consecutive recognition failures, classify the security threat level according to the number of consecutive recognition failures, and issue a security warning;

[0105] According to the set threshold of consecutive identification failures, the security threat level can be divided into three levels:

[0106] If the number of consecutive recognition failures is 1-2 times, a low threat signal is issued;

[0107] If the number of consecutive identification failures is 3-4, a medium threat signal is issued;

[0108] If the number of consecutive identification failures is 5 or more, a high threat signal will be issued.

[0109] Specifically, setting a threshold for the number of consecutive recognition failures and classifying the security threat level and issuing security warnings according to the situation can enhance the security of the system and also provide early warning of potential security risks so that corresponding measures can be taken to prevent and respond.

[0110] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An attendance punch-in system based on image recognition, characterized in that: Includes the following modules: Image acquisition module: extracts the facial image and attendance timestamp data of the attendance personnel through a high-resolution camera, and transmits the data to the image recognition module; Image recognition module: perform image preprocessing on the extracted facial images of attendance personnel and attendance timestamp data and extract key feature points of facial images, obtain comprehensive ratio calculation values ​​through the pixel ratio values ​​of eyes and nose, nose and mouth, and eyes and mouth to analyze and identify the extracted data; Image storage module: stores the facial images of the registered attendance personnel, obtains the comprehensive ratio storage calculation value set {A1, A2, A3, ..., An} according to the pixel ratio values ​​of the eyes, nose, nose and mouth, and eyes and mouth of the registered attendance personnel, and stores the set in the database; Image matching module: Based on the comprehensive feature ratio coefficient obtained by the comprehensive weight difference between the extracted facial feature key points and the stored facial feature key points, the comprehensive feature fitness value is obtained through the comprehensive ratio calculation value, the comprehensive ratio storage calculation value and the comprehensive feature ratio coefficient, and compared with the preset threshold to determine whether the match is successful; Image security module: Set the threshold for the number of consecutive recognition failures, classify the security threat level according to the number of consecutive recognition failures, and issue a security warning.

2. According to the image recognition-based attendance punching system of claim 1, it is characterized in that: The image acquisition method includes the following steps: extracting the facial image of the attendance personnel and the attendance timestamp data through a high-resolution camera: The facial image data of the attendance personnel is obtained through a high-resolution camera. The system records the current timestamp as the attendance time in real time and transmits the facial image and attendance time data of the attendance personnel to the image recognition module.

3. The attendance punching system based on image recognition according to claim 1 is characterized in that: The image recognition module performs image preprocessing on the extracted facial image and attendance time data of the attendance personnel, including image grayscale and histogram equalization, including the following steps: S201: receiving the facial image and attendance time data of the attendance personnel from the image acquisition module, and converting the color image into a grayscale image through the cvtColor function in the OpenCV cross-platform computer vision and machine learning software library; S202: Perform histogram equalization on the grayscale image using the equalizeHist function in the OpenCV cross-platform computer vision and machine learning software library.

4. The attendance punching system based on image recognition according to claim 1 is characterized in that: The extraction and detection of key feature points of the facial image comprises the following steps: By combining OpenCV and Dli ib library, the key facial feature points of the attendance personnel, including eyes, nose and mouth, are detected. According to the extracted facial nose key point coordinates A (x1nose, y1nose, z1nose) and B (x2nose, y2nose, y3nose), facial eye key point coordinates C (x1eye, y1eye, z1eye) and D (x2eye, y2eye, z2eye), facial mouth key point coordinates E (x1mouth, y1mouth, z1mouth) and F (x2mouth, y2mouth, z2mouth), origin coordinates, the pixel point ratio values ​​of the eyes and nose, nose and mouth, and eyes and mouth are calculated, and different weights are assigned to the ratio values ​​to construct a comprehensive ratio calculation value; According to the ratio of the distance between the facial nose key point coordinates A (x1nose, y1nose, z1nose) and B (x2nose, y2nose, y3nose) and the average length of the distance between the two points AB and the origin coordinates, combined with the distance between the facial eye key point coordinates C (x1eye, y1eye, z1eye) and D (x2eye, y2eye, z2eye) and the average length of the distance between the two points CD and the origin coordinates, the eye-nose pixel ratio value calculation formula is obtained by comparing the two: Among them, O nose L is the average length of the distance between points A and B and the origin of the coordinate system. nose is the ratio of the distance between the two key points AB of the nose to the average length, O eye L is the average length of the distance between points C and D and the origin of the coordinate system. eye is the ratio of the distance between the eye key point CD and the average value, R, G and B are the red, green and blue points of the color image, r1 is the calculated value of the ratio of eye and nose pixels; According to the ratio of the distance between the facial nose key point coordinates A (x1nose, y1nose, z1nose) and B (x2nose, y2nose, y3nose) and the average length of the distance between the two points AB and the origin coordinates, combined with the distance between the facial mouth key point coordinates E (x1mouth, y1mouth, z1mouth) and F (x2mouth, y2mouth, z2mouth) and the average length of the distance between the two points EF and the origin coordinates, the nose and mouth pixel ratio value calculation formula is obtained by comparing the two: Among them, O mouth L is the average length of the distance between points E and F and the origin of the coordinate system. mouth is the ratio of the distance between the two key points EF of the mouth to the average length, and r2 is the calculated value of the ratio of the pixels of the nose and mouth; According to the ratio of the distance between the facial eye key point coordinates C (x1eye, y1eye, z1eye) and D (x2eye, y2eye, z2eye) and the average length of the distance between CD and the origin coordinates, combined with the distance between the facial mouth key point coordinates E (x1mouth, y1mouth, z1mouth) and F (x2mouth, y2mouth, z2mouth) and the average length of the distance between EF and the origin coordinates, the eye-mouth pixel ratio value calculation formula is obtained by comparing the two: Among them, L eye L is the distance between the eye key points CD. mouth is the distance between the mouth key points EF, and r3 is the calculated value of the eye-mouth pixel ratio; The comprehensive ratio calculation formula is obtained by assigning different weights to the pixel ratio values ​​of eyes and nose, nose and mouth, and eyes and mouth: a=(r1×W1+r2×W2+r3×W3)×100% Among them, a is the calculated value of the comprehensive ratio, r1 is the calculated value of the ratio of eye and nose pixels, r2 is the calculated value of the ratio of nose and mouth pixels, r3 is the calculated value of the ratio of eye and mouth pixels, W1, W2, and W3 are the corresponding weights, W1 is 30%, W2 is 20%, and W3 is 50%.

5. The attendance punching system based on image recognition according to claim 4 is characterized in that: The image storage module stores the facial features of the registered attendance personnel, including the following steps: According to the key point ratio values ​​of the eyes and nose, nose and mouth, and eyes and mouth of the registered attendance personnel, a comprehensive ratio storage calculation value set {A1, A2, A3, ..., An} is obtained and stored in the database; The comprehensive ratio storage calculation formula is obtained through the stored key point ratio values ​​of eyes and nose, nose and mouth, and eyes and mouth: A i =(R1×W1+R2×W2+R3×W3)×100% Among them, each A i Represents the calculated value of the comprehensive ratio storage of the i-th registered attendance personnel, R1 is the calculated value of the eye-nose pixel ratio, R2 is the calculated value of the nose-mouth pixel ratio, R3 is the calculated value of the eye-mouth pixel ratio, i∈(1, 2, ..., n), n is the total number of attendance personnel, W1, W2, W3 are the corresponding weights, W1 is 30%, W2 is 20%, and W3 is 50%.

6. The attendance punching system based on image recognition according to claim 5 is characterized in that: The image matching module obtains a comprehensive feature ratio coefficient based on the comprehensive weight difference between the extracted facial feature key points and the stored facial feature key points, comprising the following steps: The calculation formula for comprehensive feature fitness is obtained based on the comprehensive ratio calculation value and the comprehensive ratio storage calculation value: Among them, α is the comprehensive characteristic proportional coefficient, r recognized is the extracted facial feature key point vector, R stored is the stored facial feature key point vector, R1, R2, R3 are the stored calculated values ​​of eye-nose, nose-mouth and eye-mouth pixel ratios, r1, r2, r3 are the calculated values ​​of eye-nose, nose-mouth and eye-mouth pixel ratios, W1, W2, W3 are the corresponding weights, W1 is 30%, W2 is 20%, and W3 is 50%.

7. The attendance punching system based on image recognition according to claim 6 is characterized in that: The comprehensive feature adaptation value is obtained by the comprehensive ratio calculation value, the comprehensive ratio storage calculation value and the comprehensive feature ratio coefficient, and compared with the preset threshold to determine whether the match is successful, including the following steps: Among them, α is the comprehensive feature ratio coefficient, a is the comprehensive ratio calculation value, Ai is the comprehensive ratio storage calculation value of the registered attendance personnel, and P is the comprehensive feature fitness calculation value; If P ≥ T, the match is successful and the attendance punch-in is recorded successfully; If P < T, the match fails, and a security threat warning is issued according to the set number of consecutive failures; Among them, P is the calculated value of comprehensive feature fitness, and T is the preset threshold.

8. The attendance punching system based on image recognition according to claim 7 is characterized in that: The image security module divides the security threat level according to the number of consecutive recognition failures, including the following steps: Set a threshold for the number of consecutive recognition failures, classify the security threat level according to the number of consecutive recognition failures, and issue a security warning; According to the set threshold of consecutive identification failures, the security threat level can be divided into three levels: If the number of consecutive recognition failures is 1-2 times, a low threat signal is issued; If the number of consecutive identification failures is 3-4, a medium threat signal is issued; If the number of consecutive identification failures is 5 or more, a high threat signal will be issued.

Images