Labor service real-name system management method and system and storage medium

Abstract

The invention discloses a labor service real-name system management method and system and a storage medium, and belongs to the technical field of labor service management. The method comprises the following steps: a labor service personnel information acquisition stage: acquiring human face information of labor service personnel and associating the human face information with identity information of the labor service personnel; in the labor service personnel management stage, real-time images are collected and face recognition is carried out to obtain to-be-recognized face features when labor service personnel carry out attendance checking or go in and out of a preset area, and then the to-be-recognized face features are compared with the face information of the labor service personnel to obtain a face comparison result. And displaying the on-duty number of people and the identity information of the labor service personnel in real time according to the face comparison result, and deploying the labor service personnel according to the on-duty number of people displayed in real time, the identity information of the labor service personnel and the personnel demand of the preset area. Through modern information technologies such as a face recognition technology and the like, enterprises are helped to realize informatization, datamation and intelligentization of labor service relationships, and the problems of labor service personnel in-out safety management, labor service data missing and labor service personnel salary shortage caused by poor real-time performance are solved.

Description

Technical Field

[0001] The present invention relates to the field of labor management technology, and in particular to a labor real-name management method, system and storage medium. Background Art

[0002] At present, the construction area has a complex environment, a variety of personnel, a large number of equipment and materials, and an overall lack of effective management and control. The registration of labor personnel information is chaotic and incomplete, which can easily lead to labor disputes. The main problems are as follows: 1. Personnel management is difficult: there are many types of work, a variety of workers, and high mobility. Labor registration is complicated and it is difficult to organize and retain information; 2. Project management is difficult: it is difficult to obtain real-time information on workers' entry and exit, attendance is difficult to control, and information is difficult to unify; 3. Lack of effective supervision: construction site projects are scattered, data information is lagging, and data information is difficult to unify and remotely supervise in a timely manner, resulting in the failure to pay labor personnel in time. Summary of the invention

[0003] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide a labor real-name management method, system and storage medium.

[0004] The object of the present invention is achieved through the following technical solutions: The first aspect of the present invention provides: a labor real-name management method, comprising the following steps: During the labor personnel information collection stage, the labor personnel's facial information is collected and associated with the labor personnel's identity information; During the labor personnel management stage, when labor personnel conduct attendance or enter and exit the preset area, real-time images are collected and facial recognition is performed to obtain facial features to be identified. The facial features to be identified are then compared with the labor personnel's facial information to obtain facial comparison results. The number of people on duty and the labor personnel's identity information are displayed in real time based on the facial comparison results, and labor personnel are deployed based on the real-time display of the number of people on duty, the labor personnel's identity information and the personnel needs of the preset area.

[0005] Preferably, the face recognition comprises the following steps: Performing face detection to obtain a real-time face position image in the real-time image, and then using a neural network algorithm to detect the real-time face position image to obtain real-time face key points; Performing affine transformation on the real-time face position image according to the real-time face key points to obtain a corrected face image; Input the corrected face image into the trained feature extraction model to obtain the feature vector of the face to be identified; The facial information of the labor personnel is converted into a labeled labor personnel facial feature vector, and then the cosine similarity is calculated between the facial feature vector of the person to be identified and the facial feature vector of each labor personnel. For a facial feature vector to be identified, the facial feature vector of the labor personnel with the largest cosine similarity is selected as a group of facial comparison results.

[0006] Preferably, the neural network algorithm is an MTCNN algorithm, and the use of the neural network algorithm to detect the real-time face position image to obtain the real-time face key points comprises the following steps: Scale the real-time face position image to different sizes to build an image pyramid; Input each layer of the image in the image pyramid into the P-Net network to generate candidate windows, and filter out the first face candidate window based on the probability threshold; The first face candidate window is input into the R-Net network, and the position of the first face candidate window is adjusted by re-screening and bounding box regression to obtain the second face candidate window; The second face candidate window is input into the O-Net network for final screening and bounding box regression to obtain real-time face key points.

[0007] Preferably, the input size of the P-Net network is 12×12, and the overlapping candidate windows are removed by non-maximum suppression before the first face candidate window is screened out according to the probability threshold. The input size of the R-Net network is 24×24, and the input size of the O-Net network is 48×48.

[0008] Preferably, the real-time facial key points include at least: a real-time left eye center point, a real-time right eye center point, a real-time nose tip center point, a real-time left mouth corner point and a real-time right mouth corner point.

[0009] Preferably, the step of performing face calibration on the real-time face position image by performing affine transformation according to the real-time face key points to obtain a corrected face image comprises the following steps: Get the reference face key points, and use the coordinates of the real-time face key points and the reference face key points to calculate the affine transformation parameter vector through the optimization method a ij , A× a ij =b, where A is the matrix consisting of the difference between the coordinates of the facial key points and the coordinates of the reference facial key points, and b is the vector consisting of the coordinates of the reference facial key points; According to the affine transformation parameter vector a ij Construct the affine transformation matrix M: Among them, a 11 is the first affine transformation parameter vector, a 12 is the second affine transformation parameter vector, a 13 is the third affine transformation parameter vector, a 21 is the fourth affine transformation parameter vector, a 22is the fifth affine transformation parameter vector, a 23 is the sixth affine transformation parameter vector; According to the affine transformation matrix M, each pixel point in the real-time face position image is mapped to a new position to obtain a mapped face position image, and then the mapped face position image is adjusted to a preset size and cropped to obtain a corrected face image.

[0010] Preferably, the reference facial key points include at least: a reference left eye center point, a reference right eye center point, a reference nose tip center point, a reference left mouth corner point and a reference right mouth corner point, and the optimization method is a least squares method.

[0011] Preferably, it also includes a liveness recognition stage, and the real-time face key points include at least: the first eye key point P 1. Second Eye Key Points P 2. Key points of the third eye P 3. The fourth eye key points P 4. Key points of the fifth eye P 5. Sixth Eye Key Points P 6; The first eye key point P 1. Second Eye Key Points P 2. Key points of the third eye P 3. The fourth eye key points P 4. Key points of the fifth eye P 5. Sixth Eye Key Points P 6 is the key point collected for the same eye; the first eye key point P 1 is collected at the left eye corner; the fourth eye key point P 4 is collected from the right eye; the second eye key point P 2 and the third eye key points P 3 collected on the upper eyelid, and the second eye key point P 2 and the third eye key points P 3 is the first preset distance; the fifth eye key point P 5 and 6th Eye Key Points P The spacing of 6 is the second preset distance; The living body recognition stage includes the following steps, calculating the real-time eye aspect ratio EAR of each frame of the same real-time face position image within a preset number of frames: , When the real-time eye aspect ratio EAR of at least one frame within the preset number of frames is lower than the preset threshold, it is determined that the laborer corresponding to the real-time face position image has performed a blinking action; When performing liveness recognition, the displacement of real-time facial key points is used to determine whether the target laborer has shaken his head to the left, to the right, or blinked. Each time liveness recognition is performed, the recognition order, number of left-shaking head movements, and blinking movements, as well as the number of movements and the time of movement are randomly set. When the corresponding movement is not completed within the movement time or the movement is wrong, the liveness recognition is judged to have failed. When the number of liveness recognition failures exceeds the preset number, the target laborer is judged to be at risk.

[0012] The second aspect of the present invention provides: a labor real-name system management system, used to implement any of the above-mentioned labor real-name system management methods, comprising: The labor personnel information collection module is used to collect the labor personnel's facial information and associate it with the labor personnel's identity information; The labor personnel management module is used to collect real-time images and perform face recognition to obtain facial features to be identified when the labor personnel are checking in or entering and exiting the preset area, and then use the facial features to be identified to compare with the labor personnel's facial information to obtain facial comparison results. According to the facial comparison results, the number of people on duty and the labor personnel's identity information are displayed in real time, and the labor personnel are deployed based on the real-time display of the number of people on duty, the labor personnel's identity information and the personnel needs of the preset area.

[0013] The third aspect of the present invention provides: a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by a processor, any of the above-mentioned labor real-name management methods is implemented.

[0014] The beneficial effects of the present invention are: 1) Through information system platforms, Internet of Things hardware equipment, face recognition, liveness recognition and other modern information technologies, we help enterprises realize the informatization, dataization and intelligence of labor relations, and solve the problems of labor personnel's entry and exit management, safety management, and labor data missing and poor real-time performance, resulting in wage arrears for labor personnel.

[0015] 2) It can accurately grasp the attendance of workers and the status of various personnel on the job, realize real-time dynamic management and safety supervision of labor personnel on the construction site, and improve the level of enterprise information management.

[0016] 3) It can monitor the attendance of key responsible persons and understand the real-time status of personnel online, which can help solve common behaviors in the construction industry such as subcontracting, subcontracting, illegal subcontracting, etc., and control the frequency of accidents.

[0017] 4) Through face recognition attendance management, labor management is more accurate and efficient. Workers can clock in and out using their mobile phones, which makes the clocking in and out faster. The number of people on duty and their identity information are displayed in real time. Management can easily grasp the actual number of people entering the construction site and the status of the laborers, which makes it convenient to reasonably deploy the laborers and greatly saves management and coordination costs.

[0018] 5) Real-name management allows workers’ identities to be verified, workers’ attendance is electronically recorded, and wage calculation and payment form a chain of evidence, which can effectively avoid labor disputes, protect the legitimate rights and interests of labor personnel and enterprises, and promote the harmonious development of the construction environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Provide a flow chart for the real-name management method for labor; Figure 2 This is the attendance flow chart. DETAILED DESCRIPTION

[0020] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0021] See also Figure 1-Figure 2 The first aspect of the present invention provides: a labor real-name management method, comprising the following steps: During the labor personnel information collection stage, the labor personnel's facial information is collected and associated with the labor personnel's identity information; During the labor personnel management stage, when labor personnel conduct attendance or enter and exit the preset area, real-time images are collected and facial recognition is performed to obtain facial features to be identified. The facial features to be identified are then compared with the labor personnel's facial information to obtain facial comparison results. The number of people on duty and the labor personnel's identity information are displayed in real time based on the facial comparison results, and labor personnel are deployed based on the real-time display of the number of people on duty, the labor personnel's identity information and the personnel needs of the preset area.

[0022] In this embodiment, a three-level management method is used to manage labor companies, labor teams, and labor personnel. The collected facial information of labor personnel is matched with on-site photos to create a facial database. At the same time, the basic information of labor personnel is enriched to match the corresponding faces and standardize personnel information. Through the integration of identity and ID card, facial recognition and other equipment, real-name management is carried out on every employee on the construction site, standardizing the site employment management, labor personnel attendance management, and access management.

[0023] In some embodiments, the face recognition comprises the following steps: Performing face detection to obtain a real-time face position image in the real-time image, and then using a neural network algorithm to detect the real-time face position image to obtain real-time face key points; Performing affine transformation on the real-time face position image according to the real-time face key points to obtain a corrected face image; Input the corrected face image into the trained feature extraction model to obtain the feature vector of the face to be identified; The facial information of the labor personnel is converted into a labeled labor personnel facial feature vector, and then the cosine similarity is calculated between the facial feature vector of the person to be identified and the facial feature vector of each labor personnel. For a facial feature vector to be identified, the facial feature vector of the labor personnel with the largest cosine similarity is selected as a group of facial comparison results.

[0024] In some embodiments, the neural network algorithm is an MTCNN algorithm, and using the neural network algorithm to detect the real-time face position image to obtain the real-time face key points includes the following steps: Scale the real-time face position image to different sizes to build an image pyramid; Input each layer of the image in the image pyramid into the P-Net network to generate candidate windows, and filter out the first face candidate window based on the probability threshold; The first face candidate window is input into the R-Net network, and the position of the first face candidate window is adjusted by re-screening and bounding box regression to obtain the second face candidate window; The second face candidate window is input into the O-Net network for final screening and bounding box regression to obtain real-time face key points.

[0025] MTCNN is an efficient multi-task cascade convolutional neural network for face detection and key point location. It achieves accurate face detection through three-stage networks (P-Net, R-Net, and O-Net). Face recognition through MTCNN first requires the preparation of a large amount of labeled data, which usually comes from face datasets such as WIDER FACE, LFW, etc. These images contain labeled information of face frames, which need to be converted into a form suitable for network input through data preprocessing. In order to train an efficient MTCNN model, some training techniques can be used, such as positive and negative sample balance, data enhancement, and joint training.

[0026] In some embodiments, the input size of the P-Net network is 12×12, and the overlapping candidate windows are removed by non-maximum suppression before the first face candidate window is screened out according to the probability threshold. The input size of the R-Net network is 24×24, and the input size of the O-Net network is 48×48.

[0027] In this embodiment, P-Net is a fully convolutional network with an input size of 12x12, which is used to quickly generate a large number of face candidate frames; the input size of R-Net is 24x24, which is used to refine the first face candidate frame generated by P-Net, and perform more detailed classification and regression on each first face candidate frame to improve the accuracy of face detection; O-Net is the largest network in MTCNN, with an input size of 48x48, and performs final classification and regression on the second face candidate frame generated by R-Net to obtain the most accurate face detection results and provide real-time face key point positions.

[0028] In some embodiments, the real-time facial key points include at least: a real-time left eye center point, a real-time right eye center point, a real-time nose tip center point, a real-time left mouth corner point, and a real-time right mouth corner point.

[0029] In some embodiments, the method of performing face calibration on the real-time face position image according to the real-time face key points by performing affine transformation to obtain a corrected face image comprises the following steps: Get the reference face key points, and use the coordinates of the real-time face key points and the reference face key points to calculate the affine transformation parameter vector through the optimization method a ij , A× a ij =b, where A is the matrix consisting of the difference between the coordinates of the facial key points and the coordinates of the reference facial key points, and b is the vector consisting of the coordinates of the reference facial key points; According to the affine transformation parameter vector a ij Construct the affine transformation matrix M: Among them, a 11 is the first affine transformation parameter vector, a 12 is the second affine transformation parameter vector, a 13 is the third affine transformation parameter vector, a 21 is the fourth affine transformation parameter vector, a 22 is the fifth affine transformation parameter vector, a 23 is the sixth affine transformation parameter vector; According to the affine transformation matrix M, each pixel point in the real-time face position image is mapped to a new position to obtain a mapped face position image, and then the mapped face position image is adjusted to a preset size and cropped to obtain a corrected face image.

[0030] In this embodiment, the preset size is a size pre-set according to the actual working environment, working requirements, etc., and is not limited to a specific value or range of values, as long as the purpose of the present invention can be achieved. The reference face key points are key points used to define the standard face posture. Usually, a standard face image is selected, and its key points are evenly distributed, such as a frontal face image. The affine transformation matrix M of the real-time face key points (5 landmarks) and the reference face key points (reference landmark) is estimated by using the estimateAffine2D() function of OpenCV, and then the real-time face position image is corrected according to the affine transformation matrix M, which can effectively improve the effect of face recognition. Through various image processing libraries, each pixel in the real-time face position image can be mapped to a new position according to the affine transformation matrix M to obtain a mapped face position image, such as OpenCV, PIL, etc., which are used to apply affine transformation and handle image boundaries and other issues. After obtaining the mapped face position image, some post-processing steps may be required, such as resizing, cropping to remove excess background, etc., to ensure that the face is centered in the image and of the right size.

[0031] In some embodiments, the reference facial key points include at least: a reference left eye center point, a reference right eye center point, a reference nose tip center point, a reference left mouth corner point and a reference right mouth corner point, and the optimization method is the least squares method.

[0032] In some embodiments, a liveness recognition stage is also included, and the real-time face key points include at least: a first eye key point P 1. Second Eye Key Points P 2. Key points of the third eye P 3. The fourth eye key points P 4. Key points of the fifth eye P 5. Sixth Eye Key Points P 6; The first eye key point P 1. Second Eye Key Points P 2. Key points of the third eye P 3. The fourth eye key points P 4. Key points of the fifth eye P 5. Sixth Eye Key Points P 6 is the key point collected for the same eye; the first eye key point P 1 is collected at the left eye corner; the fourth eye key point P 4 is collected from the right eye; the second eye key point P 2 and the third eye key points P 3 collected on the upper eyelid, and the second eye key point P 2 and the third eye key points P3 is the first preset distance; the fifth eye key point P 5 and 6th Eye Key Points P The spacing of 6 is the second preset distance; The living body recognition stage includes the following steps, calculating the real-time eye aspect ratio EAR of each frame of the same real-time face position image within a preset number of frames: , When the real-time eye aspect ratio EAR of at least one frame within the preset number of frames is lower than the preset threshold, it is determined that the laborer corresponding to the real-time face position image has performed a blinking action; When performing liveness recognition, the displacement of real-time facial key points is used to determine whether the target laborer has shaken his head to the left, to the right, or blinked. Each time liveness recognition is performed, the recognition order, number of left-shaking head movements, and blinking movements, as well as the number of movements and the time of movement are randomly set. When the corresponding movement is not completed within the movement time or the movement is wrong, the liveness recognition is judged to have failed. When the number of liveness recognition failures exceeds the preset number, the target laborer is judged to be at risk.

[0033] In this embodiment, the first preset distance, the second preset distance, the preset number of frames, the preset threshold, the number of actions, the action time, and the preset number of times are preset values ​​according to the actual working environment and working requirements, and are not limited to a specific value or value range. The liveness recognition stage is to draw a conclusion on whether it is a live body by judging whether the same real-time face position image within the preset number of frames is blinking. The basic principle of judging blinking: judging whether the human eye has a blinking action based on the change of the real-time eye aspect ratio EAR (EyeAspect Ratio). When the human eye is open, the EAR fluctuates around a certain value, and when the human eye is closed, the EAR drops rapidly, and in theory it will be close to zero. Therefore, we can assume that when the EAR is lower than the preset threshold, the eye is in a closed state, and as long as there is one frame in the preset number of frames where the EAR is lower than the preset threshold, it means that a blinking action has been performed. The blinking speed is relatively fast, and generally 1 to 3 frames complete the blinking action. Since the user needs to complete various actions, paper or electronic photos basically cannot pass this liveness detection. However, for videos, attackers may use pre-recorded videos that complete a certain sequence of actions to deceive the system. For this situation, the countermeasures of the present invention are as follows: the system requires the user to complete the left shaking head action, the right shaking head action, and the blinking action (the mouth opening action can also be added, but the facial key points required for the mouth opening action to be identified will increase, and the amount of data for identification and calculation will also increase accordingly, while the left shaking head action, the right shaking head action, and the blinking action can all be judged only by six eye key points, which greatly reduces the amount of data for identification and calculation, improves the recognition efficiency, and reduces the requirements for the system's computing power), the system randomly scrambles the recognition order of the above actions, and the number of blinking actions (mouth opening actions) is also a random number. In the above manner, when the user performs each interactive liveness detection, the scheme that needs to be completed is completely different, and the number of blinking actions (mouth opening actions) completed is also different. When the user fails to complete the detection within the specified time of the system, it is automatically judged that the liveness detection fails. When the user fails to log in for more than 3 times, the system will judge that the current user is at risk and lock the currently logged in user. The locked user needs to be unlocked by the administrator through the administrator system. In the above manner, the method is also capable of resisting and blocking video deception attacks.

[0034] The second aspect of the present invention provides: a labor real-name system management system, used to implement any of the above-mentioned labor real-name system management methods, comprising: The labor personnel information collection module is used to collect the labor personnel's facial information and associate it with the labor personnel's identity information; The labor personnel management module is used to collect real-time images and perform face recognition to obtain facial features to be identified when the labor personnel are checking in or entering and exiting the preset area, and then use the facial features to be identified to compare with the labor personnel's facial information to obtain facial comparison results. According to the facial comparison results, the number of people on duty and the labor personnel's identity information are displayed in real time, and the labor personnel are deployed based on the real-time display of the number of people on duty, the labor personnel's identity information and the personnel needs of the preset area.

[0035] In this embodiment, the labor real-name management system also combines intelligent hardware devices such as smart helmets and uses the Internet of Things positioning module to effectively control and remotely monitor personnel entering and leaving the construction area, attendance, activity trajectories, etc. It can effectively strengthen construction site safety management, reduce the frequency of accidents, eliminate various illegal operations and uncivilized construction, strengthen labor personnel management, avoid labor disputes, protect the legitimate rights and interests of labor personnel and enterprises, and promote the harmonious development of the construction environment.

[0036] The third aspect of the present invention provides: a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by a processor, any of the above-mentioned labor real-name management methods is implemented.

[0037] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A labor real-name management method, characterized by: The following steps are involved: During the labor personnel information collection stage, the labor personnel's facial information is collected and associated with the labor personnel's identity information; During the labor personnel management stage, when labor personnel conduct attendance or enter and exit the preset area, real-time images are collected and facial recognition is performed to obtain facial features to be identified. The facial features to be identified are then compared with the labor personnel's facial information to obtain facial comparison results. The number of people on duty and the labor personnel's identity information are displayed in real time based on the facial comparison results, and labor personnel are deployed based on the real-time display of the number of people on duty, the labor personnel's identity information and the personnel needs of the preset area.

2. The labor real-name management method according to claim 1 is characterized by: The face recognition comprises the following steps: Performing face detection to obtain a real-time face position image in the real-time image, and then using a neural network algorithm to detect the real-time face position image to obtain real-time face key points; Performing affine transformation on the real-time face position image according to the real-time face key points to obtain a corrected face image; Input the corrected face image into the trained feature extraction model to obtain the feature vector of the face to be identified; The facial information of the labor personnel is converted into a labeled labor personnel facial feature vector, and then the cosine similarity is calculated between the facial feature vector of the person to be identified and the facial feature vector of each labor personnel. For a facial feature vector to be identified, the facial feature vector of the labor personnel with the largest cosine similarity is selected as a group of facial comparison results.

3. The labor real-name management method according to claim 2 is characterized by: The neural network algorithm is an MTCNN algorithm, and the use of the neural network algorithm to detect the real-time face position image to obtain the real-time face key points includes the following steps: Scale the real-time face position image to different sizes to build an image pyramid; Input each layer of the image in the image pyramid into the P-Net network to generate candidate windows, and filter out the first face candidate window based on the probability threshold; The first face candidate window is input into the R-Net network, and the position of the first face candidate window is adjusted by re-screening and bounding box regression to obtain the second face candidate window; The second face candidate window is input into the O-Net network for final screening and bounding box regression to obtain real-time face key points.

4. The labor real-name management method according to claim 3 is characterized by: The input size of the P-Net network is 12×12. Before the first face candidate window is screened out according to the probability threshold, the overlapping candidate windows are removed by non-maximum suppression. The input size of the R-Net network is 24×24, and the input size of the O-Net network is 48×48.

5. The labor real-name management method according to any one of claims 2 to 4, characterized in that: The real-time key points of the face include at least: a real-time left eye center point, a real-time right eye center point, a real-time nose tip center point, a real-time left mouth corner point and a real-time right mouth corner point.

6. The labor real-name management method according to claim 2 is characterized in that: The method of performing face calibration on the real-time face position image by performing affine transformation according to the real-time face key points to obtain the corrected face image comprises the following steps: Get the reference face key points, and use the coordinates of the real-time face key points and the reference face key points to calculate the affine transformation parameter vector through the optimization method a ij , A× a ij =b, where A is the matrix consisting of the difference between the coordinates of the facial key points and the coordinates of the reference facial key points, and b is the vector consisting of the coordinates of the reference facial key points; According to the affine transformation parameter vector a ij Construct the affine transformation matrix M: Among them, a 11 is the first affine transformation parameter vector, a 12 is the second affine transformation parameter vector, a 13 is the third affine transformation parameter vector, a 21 is the fourth affine transformation parameter vector, a 22 is the fifth affine transformation parameter vector, a 23 is the sixth affine transformation parameter vector; According to the affine transformation matrix M, each pixel point in the real-time face position image is mapped to a new position to obtain a mapped face position image, and then the mapped face position image is adjusted to a preset size and cropped to obtain a corrected face image.

7. The labor real-name management method according to claim 6 is characterized in that: The reference facial key points include at least: a reference left eye center point, a reference right eye center point, a reference nose tip center point, a reference left mouth corner point and a reference right mouth corner point, and the optimization method is a least squares method.

8. The labor real-name management method according to claim 2 is characterized in that: It also includes a liveness recognition stage, wherein the real-time face key points include at least: the first eye key point P 1. Second Eye Key Points P 2. Key points of the third eye P 3. The fourth eye key points P 4. Key points of the fifth eye P 5. Sixth Eye Key Points P 6; The first eye key point P 1. Second Eye Key Points P 2. Key points of the third eye P 3. The fourth eye key points P 4. Key points of the fifth eye P 5. Sixth Eye Key Points P 6 is the key point collected for the same eye; the first eye key point P 1 is collected at the left eye corner; the fourth eye key point P 4 is collected from the right eye; the second eye key point P 2 and the third eye key points P 3 collected on the upper eyelid, and the second eye key point P 2 and the third eye key points P 3 is the first preset distance; the fifth eye key point P 5 and 6th Eye Key Points P The spacing of 6 is the second preset distance; The living body recognition stage includes the following steps, calculating the real-time eye aspect ratio EAR of each frame of the same real-time face position image within a preset number of frames: ， When the real-time eye aspect ratio EAR of at least one frame within the preset number of frames is lower than the preset threshold, it is determined that the laborer corresponding to the real-time face position image has performed a blinking action; When performing liveness recognition, the displacement of real-time facial key points is used to determine whether the target laborer has shaken his head to the left, to the right, or blinked. Each time liveness recognition is performed, the recognition order, number of left-shaking head movements, right-shaking head movements, and blinking movements are randomly set. When the corresponding action is not completed within the action time or the action is wrong, the liveness recognition is judged to have failed. When the number of liveness recognition failures exceeds the preset number, the target laborer is judged to be at risk.

9. A labor real-name management system, characterized by: A method for implementing the labor real-name management method as claimed in any one of claims 1 to 8, comprising: A labor personnel information collection module is used to collect the labor personnel's facial information and associate it with the labor personnel's identity information; The labor personnel management module is used to collect real-time images and perform face recognition to obtain facial features to be identified when the labor personnel are checking in or entering and exiting the preset area, and then use the facial features to be identified to compare with the labor personnel's facial information to obtain facial comparison results. According to the facial comparison results, the number of people on duty and the labor personnel's identity information are displayed in real time, and the labor personnel are deployed based on the real-time display of the number of people on duty, the labor personnel's identity information and the personnel needs of the preset area.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by the processor, the labor real-name management method as described in any one of claims 1 to 8 is implemented.

Images