Automatic invigilation device for sports test and unmanned automatic invigilation method

Abstract

The invention relates to an automatic invigilation device for sports test and an unmanned automatic invigilation method, and the invigilation method comprises the steps: transmitting the identity information of an examinee to be tested to an examinee information module through the automatic invigilation device for sports test; an information acquisition area of the image acquisition module is adjusted to a test area or sports facilities for testing, identity information of an examinee is read through the card reader, facial information of the examinee is acquired through the image acquisition module, facial feature information of the examinee is extracted through the facial feature extraction module, and after identity recognition is performed on the examinee, the examinee can be tested. The information and the information acquisition time are stored in an examinee information module; the image acquisition module is used for recording the actions of the test items performed by the examinees in the whole process; the sports test automatic invigilation software is used for performing standard judgment on actions of test items performed by examinees, performing score measurement and statistics on the standard actions, and giving test scores of the examinees, so that unmanned whole-course automatic invigilation in the sports test process is realized.

Description

Technical Field

[0001] The invention relates to a sports project testing device, in particular to an automatic proctoring device for sports project testing and an unmanned automatic proctoring method. Background Art

[0002] The existing physical test items of various schools mainly include the following three categories: items with number as the result (such as pull-ups), items with range as the result (such as standing long jump), and items with process as the result (such as dribbling over the bar). The physical test methods of schools mainly include manual testing and assisted testing with invigilation auxiliary equipment. Manual testing is conducted on-site by the examinee, supervised by the invigilator, and the test results and grades are obtained by the invigilator. The test results of this test method are obtained by the invigilator, so there is a certain error, and when the number of participants increases, the error may be greater. The assisted test of invigilation auxiliary equipment with photoelectric sensors is conducted by the examinee using the invigilation auxiliary equipment to obtain more accurate test results and grades. The biggest disadvantage of this method of assisted testing with invigilation equipment is that it is impossible to determine whether the examinee has cheated in the exam. For example, in the pull-up test item, the examinee's pull-up action is sensed and counted by the infrared search head on the equipment. When the examinee's head is between the two infrared search heads at both ends of the horizontal bar, the infrared signal is cut off, and the equipment records the completion of a pull-up action. If the examinee does not complete the prescribed movements, but only moves his head up and down between the two infrared search heads several times, the device will count the completion of multiple pull-ups, thus resulting in cheating in the exam. Therefore, the existing invigilation auxiliary equipment cannot solve the problem of cheating in the exam.

[0003] In addition, the invigilation auxiliary equipment cannot identify the behavior of others taking the exam for you, and cannot effectively identify whether the examinee's examination movements are in compliance with the regulations during the examination, whether there is any substitute taking the exam, etc. Summary of the invention

[0004] The purpose of the present invention is to provide an automatic proctoring device and an unmanned automatic proctoring method for sports test, so as to solve the problems of cheating in sports test, such as the inability of existing proctoring auxiliary equipment to identify irregular movements, inaccurate number of times, and others taking the test for others.

[0005] The present invention is achieved in that: An automatic proctoring device for sports test, comprising: The chassis is a vertical box with rollers, used to carry the components of the device; A display touch screen is arranged on the panel of the chassis and connected to the mainboard, and is used to select test items and display test results; The card reader is arranged on the inner side of the panel of the chassis and connected to the main board, and is used to read the ID card information of the examinee; The image acquisition module is arranged on the panel of the chassis and connected to the main board, and is used to capture the facial information of the test takers and to photograph the completion of the movements of the physical test items performed by the test takers; The audio module is arranged inside the chassis and connected to the mainboard, and is used to announce the test scores of the examinees in sports events; The main board is arranged inside the chassis, and is used to receive information inputted by the display touch screen, the card reader, and the image acquisition module, and after identifying the examinee, runs the automatic proctoring software for the physical test, determines whether the actions of the test items performed by the examinee meet the standards, measures and counts the results of the actions that meet the standards, and gives the examinee's test scores; A communication module, which is arranged inside the chassis and connected to the mainboard, and is used to upload the test results to the host computer through the network; and The power supply is installed inside the chassis to provide working power for each part.

[0006] Furthermore, the automatic proctoring device for sports test of the present invention also includes: The hard disk video recorder is arranged inside the chassis and connected to the image acquisition module and the main board respectively, and is used to record and save the movements of the physical test items performed by the examinee.

[0007] Furthermore, the sports test automatic proctoring software includes: Candidate information module, used to store and call the candidate's identity information; A face feature extraction module is used to extract face features from the facial information of the examinee acquired by the image acquisition module; A face verification module, used to compare the face features extracted by the face feature extraction module with the identity information of the examinee read in by the card reader and the examinee information retrieved from the examinee information module, so as to verify the authenticity of the examinee; and The candidate's score and test process recording module is used to record the candidate's physical test item completion process, determine whether the candidate's completion actions of the test items meet the standards, measure and count the scores of the completion actions that meet the standards, and give the test scores of the test items taken by the candidate.

[0008] Furthermore, the method for judging whether the candidate's movements meet the standards is as follows: calibration points are respectively set on the human body's trunk, limbs, hands and feet, and during the test, the image acquisition module is used to track and identify the corresponding calibration points of the human body in real time, and the relative relationship and motion relationship of each calibration point are judged, and the position relationship parameters between the corresponding calibrations of the candidate during the test are calculated, and compared with the position relationship parameters between the corresponding calibrations of the human body required for the standard movements of the tested sports item, thereby obtaining a judgment result on whether the movements completed by the candidate in this test item meet the standards.

[0009] Furthermore, the set calibration points include trunk joint points, arm joint points, leg joint points, hand-related points and foot-related points; the calibrated trunk joint points include shoulder joints and hip joints; the calibrated arm joint points include wrist joints and elbow joints; the calibrated leg joint points include knee joints and ankle joints; the calibrated hand-related points include the fingertips, first joints, second joints and palm joints of each finger; the calibrated foot-related points include toes and heels.

[0010] The present invention can also be implemented like this: A method for unmanned automatic proctoring of sports test, comprising the following steps: S1. Setting the automatic proctoring device for sports test as described in any one of claims 1 to 5 to transmit the identity information of the examinee to the examinee information module; S2. Set the automatic proctoring device for sports test in the examination room, adjust the information acquisition area of ​​the image acquisition module to the test area or the sports facilities for test, select the test sports through the display touch screen, and select the test sports and their test standards according to the gender and age group of the examinee; S3, reading the candidate's identity information through the card reader, collecting the candidate's facial information through the image acquisition module, extracting the candidate's facial feature information through the facial feature extraction module, and after identifying the candidate, storing the above information and information collection time in the candidate information module; S4. Use the image acquisition module to record the entire process of the test items completed by the examinee; based on the video information of the test items completed by the examinee, combined with the automatic monitoring software for sports test running on the mainboard, the completion actions of the test items performed by the examinee are judged to be up to standard, the scores of the completion actions that meet the standards are measured and counted, and the test scores of the test items performed by the examinee are given.

[0011] Furthermore, the method for judging whether the candidate's movements meet the standards is as follows: the method for judging whether the candidate's movements meet the standards is as follows: the joints of the human trunk, limbs, hands and feet are respectively calibrated, and during the test, the image acquisition module is used to track and identify the calibrated human joints in real time, and the relative relationship and motion relationship of each joint are judged, the joint parameters of the candidate during the test are calculated and identified, and the results of whether the candidate's movements meet the standards are obtained by comparing them with the human joint parameters required for the standard movements of the tested sports.

[0012] Furthermore, the image acquisition module processes the collected test action video of the examinee by diverting the video information into multiple video information streams, corresponding to work tasks including face recognition, limb recognition, hand recognition, body recognition, multi-target body recognition, and real-time tracking of body posture recognition, and synchronizes and coordinates each video stream through a synchronizer.

[0013] The present invention processes the video captured by the camera and then divides it into different video streams. These video streams are used for tasks such as face recognition, limb recognition, hand recognition, human body recognition, multi-target human body recognition, real-time tracking of human posture recognition, etc. Whenever a new task is added, a new video stream will be automatically added; and through the coordination of the synchronizer, the video streams can be processed simultaneously and kept synchronized. The results obtained by each video stream will be stored and processed uniformly to provide a more comprehensive and accurate recognition result. As a result, the present invention can uniformly process all the information that can be obtained from the video, and different recognitions can cooperate with each other to complete complex task scheduling.

[0014] This processing method greatly improves the efficiency, convenience and cost saving of multi-tasking. Compared with the solutions on the market that require the use of multiple cameras, the matrix video splitting synchronous recognition technology of the present invention can achieve more efficient and accurate output results through a single camera. It not only provides a more convenient solution, but also provides more comprehensive and accurate technical support for recognition tasks in different application scenarios.

[0015] The present invention also has good versatility, encapsulating the intelligent recognition part as a whole, and by sending the video to be recognized to the system for processing, all available data in the video can be extracted and stored.

[0016] The recognition technology of the present invention includes but is not limited to sports events. It only needs to inform the system of the start and end of the event. The process will be processed differently according to different categories. For example, in the category with the number as the result, the system of the present invention will download the standard video from the server, and carry the key frame information mark and the key detection points of the corresponding project, automatically extract the human body features and record the key actions of the video data obtained from the camera, and compare them with the human body features and key actions extracted from the standard video. In this way, the system can determine whether the content in the video data obtained by the camera meets the standard. For other categories, the system will process the video according to different processing methods.

[0017] The present invention can automatically record the test situation and give the test score of the examinee through the automatic proctoring device for sports test, quantify various indicators in the test process by reading the examinee's limb parameters, and the result can be traced back. The test result of the examinee can be identified through video recording to reproduce the test process of the examinee; cheating in the test (such as stepping on a support or holding the bar with reverse hands in the pull-up project and hand support, inadequate movements, bent legs in the sit-up project, holding auxiliary tools, and inclined sitting test) can all be accurately captured, thereby overcoming the disadvantage that the existing auxiliary proctoring device cannot identify cheating in the test, and can realize parameterization and indexation of the sports project test process, and realize unmanned automatic proctoring throughout the process. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the external structure of the automatic invigilation device for sports test of the present invention.

[0019] Figure 2 The present invention is a schematic diagram of the internal structure of the automatic invigilation device for sports test.

[0020] Figure 3 It is a calibration diagram that sets calibration points on the human body's trunk, limbs, hands and feet.

[0021] Figure 4 It is a video streaming flow chart.

[0022] In the figure: 1. base, 2. chassis, 3. display touch screen, 4. card reader, 5. image acquisition module, 6. communication module, 7. main board, 8. audio module, 9. power supply, 10. switch, 11. second main board, 12. second image acquisition module, DETAILED DESCRIPTION

[0023] like Figure 1 , Figure 2As shown, the automatic proctoring device for sports project test of the present invention includes a chassis 2, a display touch screen 3, a card reader 4, an image acquisition module 5, an audio module 8, a main board 7, a communication module 6 and a power supply 9. The chassis 2 is a vertical box, and a roller is installed under the base 1 at the bottom. The chassis 2 is used to carry various components of the device. The display touch screen 3 is arranged on the panel of the chassis 2 and is electrically connected to the main board 7, and is used to select the test items and display the test results. The card reader 4 is arranged on the inner side of the panel of the chassis 2 and is electrically connected to the main board 7, and is used to read the ID card information of the examinee. The image acquisition module 5 is arranged on the panel of the chassis 2 and is electrically connected to the main board 7, and is used to capture the facial information of the examinee and shoot the action completion status of the sports test items performed by the examinee. The audio module 8 is arranged inside the chassis 2 and is electrically connected to the main board 7, and is used to broadcast the sports test scores of the examinee. The main board 7 is arranged inside the chassis 2, and is used to receive information input by the display touch screen 3, the card reader 4, and the image acquisition module 5. After the examinee is identified, the automatic proctoring software for the sports test is run to determine whether the actions of the test items performed by the examinee meet the standards, measure and count the results of the standard actions, and give the examinee's test results. The communication module 6 is arranged inside the chassis 2, and the antenna is arranged on the top of the chassis; the communication module 6 is electrically connected to the main board 7, and is used to upload the test results to the host computer through the network. The power supply 9 is arranged inside the chassis 2 to provide working power for each part.

[0024] A hard disk video recorder (not shown) is also provided in the chassis 2, which is connected to the image acquisition module 5 and the main board 7, and is used to record and save the movements of the physical test items performed by the examinee.

[0025] Motherboard 7 uses a high-performance 8-core processor with a main frequency of up to 2.4Ghz and a 4-core GPU. It also supports multiple display interfaces, multiple network interfaces, and a rich range of expansion interfaces (USB2.0, USB3.0, TTL, RS232, RS485, SPK, CAN, GPIO, SATA, PCle3.0x2, etc.), and supports Android and Linux systems. At the same time, the motherboard supports three recognition functions: face recognition, multi-target human body recognition, and real-time tracking of human posture recognition.

[0026] The unmanned automatic proctoring method for sports test of the present invention comprises the following steps: 1. Set up an automatic proctoring device for sports tests to transmit the identity information of candidates to the candidate information module.

[0027] 2. Set up the automatic invigilation device for physical education tests in the examination room, and adjust the information acquisition area of ​​the image acquisition module to the test area or the sports facilities for testing. Select the test sports through the touch screen, and select the test sports and their test standards based on the gender and age group of the examinee.

[0028] 3. Read the candidate's identity information through the card reader, collect the candidate's facial information through the image acquisition module, extract the candidate's facial feature information through the facial feature extraction module, and after the candidate's identity is recognized, store the above information and information collection time in the candidate information module.

[0029] 4. Use the image acquisition module to record the entire process of completing the test items performed by the candidates; based on the video information of the test items completed by the candidates, combined with the automatic monitoring software for sports tests running on the mainboard, the completion actions of the test items performed by the candidates are judged to be up to standard, the scores of the completion actions that meet the standards are measured and counted, and the test scores of the test items performed by the candidates are given.

[0030] The automatic proctoring software for sports project tests includes a candidate information module, a face feature extraction module, a face verification module, and a candidate score and test process recording module. The candidate information module is used to store and call the candidate's identity information. The face feature extraction module is used to extract face features from the candidate's facial information captured by the image acquisition module. The face verification module is used to compare the face features extracted by the face feature extraction module with the candidate's identity information read in by the card reader and the candidate information called out from the candidate information module to verify the authenticity of the candidate. The candidate score and test process recording module is used to record the action completion process of the candidate's sports test items, determine whether the completion action of the test item performed by the candidate meets the standard, measure and count the completion actions that meet the standard, and give the test scores of the test items performed by the candidate.

[0031] The method for judging whether the candidate's movements meet the standards is as follows: calibration points are set on the human body's trunk, limbs, hands and feet respectively; during the test, the image acquisition module is used to track and identify the corresponding calibration points of the human body in real time; by judging the relative relationship and motion relationship of each calibration point, the positional relationship parameters between the candidate's corresponding calibrations during the test are calculated; and by comparing them with the positional relationship parameters between the corresponding calibrations of the human body required for the standard movements of the tested sports item, a judgment result on whether the candidate's movements completed in this test item meet the standards is obtained.

[0032] like Figure 3 As shown, the human body calibration points set by the present invention include trunk joint points, arm joint points, leg joint points, hand-related points and foot-related points. The calibrated trunk joint points include shoulder joints and hip joints; the calibrated arm joint points include wrist joints and elbow joints; the calibrated leg joint points include knee joints and ankle joints; the calibrated hand-related points include the fingertips, first joints, second joints and palm joints of each finger; the calibrated foot-related points include toes and heels.

[0033] The image acquisition module processes the collected test action videos of the candidates by diverting the video information into multiple video information streams, corresponding to work tasks including face recognition, limb recognition, hand recognition, body recognition, multi-target body recognition, and real-time tracking of body posture recognition, and synchronizes and coordinates the video streams through a synchronizer.

[0034] After placing the automatic proctoring device of the present invention in a suitable position, turn on the device, and after the camera in the image acquisition module transmits the picture to the display touch screen, it will mark the test area and place the auxiliary equipment in the test area. After placing the auxiliary equipment corresponding to the physical test items in the test area, the automatic proctoring device is now placed. The automatic proctoring device reads the candidate's ID card information through the card reader 4 and performs face recognition. At this moment, the camera will record the entire process, track the candidate's location information in real time during the test, and make a record of suspected substitution of the test for the test to avoid the occurrence of substitution of others. At the same time, multi-target human body recognition is performed during the test, and real-time verification and verification are performed; whether there are multiple people in the test area, whether there are people assisting in the test, etc. A variety of methods are used to jointly prevent others from taking the test for you.

[0035] Take the physical test item - pull-ups as an example. During the test, the camera captures images to check whether the candidate's legs are bent, whether the chin is over the bar, whether the body height drops to a certain level, whether the distance between the two hands is too short, and whether the two hands are gripped in the right and wrong directions.

[0036] During the test, the examinee's human body calibration points related to pull-ups are tracked and identified in real time. By judging the relative relationship and motion relationship of each joint point / related point when the examinee completes the movement, the examinee's joint parameters during the test are calculated and identified. By comparing them with the human body joint parameters required for the standard movements of the tested sports, a judgment result is obtained as to whether the examinee's movement meets the standard.

[0037] The specific use and testing method of the automatic invigilation device of the present invention is as follows: 1. Before the exam, each candidate collects photo information and personal identity information and uploads it to the candidate information module. At the same time, the facial feature extraction module will extract the facial feature information and store it in the candidate information module.

[0038] 2. Configure the number of tests and candidate information for each candidate, as well as the various judgment parameters of the item (such as the lying distance for sit-ups, the distance between the elbow and the knee when standing up, etc.).

[0039] 3. Configure the examination room information and candidate information, associate the information with the equipment, and place the automatic invigilation device as required.

[0040] After the automatic invigilation device is powered on, the entire device is powered on through the power supply 9. The image acquisition module 5 will display the on-site information on the display touch screen 3, and will also mark the examination area and the auxiliary equipment placement area. After the auxiliary equipment of the corresponding project is placed in the test area, the device placement is completed.

[0041] Fourth, the automatic invigilation device is connected to the network through the communication module 6, and the configuration is downloaded to the local machine so that the device can perform invigilation and evaluation as required, and at the same time, after storing the examinee information, unmanned automatic invigilation is performed offline.

[0042] 5. Before normal operation, the entire examination room information and evaluation parameter information will be downloaded to the machine, and then automatically loaded when the machine is turned on, and the whole process will be recorded. If the network is connected, the examination room information will be uploaded in real time. If not, it will be stored in the machine and automatically uploaded after the network is connected.

[0043] 6. During work, multiple areas can be tested simultaneously, that is, the second mainboard is used to read and analyze the image information of the second image acquisition module in real time through the switch, analyze and record the time when the candidates enter and exit the examination area and the time when the invigilators enter and exit the examination area, and upload the above information to the candidate score and examination process recording module. At the same time, the hard disk recorder records the video process of the test in real time, and judges whether the candidates enter and exit the test area during the examination and whether the position is correct by comparing the test time and log records of the candidates in each project, combined with the candidate's position, entry and exit time, etc., so as to supervise the invigilation of the entire examination room and whether the candidates take the test as required.

[0044] 7. Multi-target human recognition The image information collected by the image acquisition module 5 is scaled and processed into the size and data structure NCHW required by the multi-target human recognition model, where "N" is the batch size, "C" is the number of channels of the feature map, "H" is the height of the feature map, and "W" is the width of the feature map; for a frame of image, N=1, C=3, H=300, and W=300. The processed image features are normalized and input to the model. The model is internally subjected to feature extraction through a deep neural network to finally form NCT, where N corresponds to several images, C represents the number of predicted human borders on a picture, and T represents the position and confidence rate of the human border. For a frame of image, the final dimensional feature N=1, C=1709 (4×4+18×18+37×37), T=5. For the first eigenvalue of T, it represents the confidence rate X. When its value X>0.5, it is considered to be a human body, otherwise it is considered not to be a human body. For the human body, the remaining 4 dimensions are the human body bounding box features. The human feature information of the current image is obtained by real-time analysis of each frame, and the 4-dimensional features are reversely transformed into the size and position of the predicted border, such as Left, Top, Width, Height (0.1, 0.2, 0.2, 0.4), which respectively mean: the distance from the upper left vertex to the left side, the distance from the upper left vertex to the upper side, the width of the predicted box, and the height of the predicted box.

[0045] 8. Real-time tracking of human posture recognition The image information collected by the image acquisition module 5 is scaled and processed into the size and data structure NCHW required by the human posture recognition model, "N" is the batch size, "C" is the number of channels of the feature map, "H" is the height of the feature map, and "W" is the width of the feature map; for a frame of image, N=1, C=3, H=300, W=300, the processed image features are normalized and input to the model, and the model is internally subjected to feature extraction through a deep neural network to finally form NCT, where N corresponds to several images, C represents the predicted human posture recognition on one image, and T represents the key point information and confidence rate of the human posture. For a frame of image, the final dimensional feature N=1, C=344 (2×2+4×4+18×18), T=163. For the first eigenvalue of T, it means that The confidence rate X, when its value X>0.5, is considered to be a human body, otherwise it is considered not a human body. For a human body, the remaining 163 dimensions are 54 key point features of the human body. The key point feature information of the current image is obtained by real-time analysis of each frame, and the 163-dimensional features are reversely transformed into the position of the predicted key points. For example, the key point information of the human body is recorded from the second position. Its meaning is shown in the following figure. Each point is represented in the order of x, y, and z; x represents the distance percentage from the left side (range value from 0-1), y represents the distance percentage from the upper side (range value from 0-1), and z represents the distance before and after the center point of the line connecting key points 56 and 47 is 0.5 (range from 0 to 1, indicating that the actual space height corresponds to the pixel ratio represented by the height in the figure for conversion), so as to obtain the coordinate information of 54 key points of the human body. For example, the coordinates of the fingertip of the right thumb are in the 2nd, 3rd, and 4th positions of the 163-dimensional (x=0.1, y=0.2, z=0.45).

[0046] If the initial image size is 1920×1080, then its conversion value is: x=0.1×1920; y=0.2×1080; z=(0.45-0.5)×1080.

[0047] IX. Single Project Examination Process After the examinee enters the test site, he / she comes to the automatic invigilation device and is scanned by the device. The card reader 4 identifies the examinee's information, the main board 7 collects the examinee's facial information through the image acquisition module 5, and uploads the facial image to the facial verification module through the communication module 6. The facial verification module calls the facial feature extraction module to obtain the examinee's facial feature information recorded in the facial information and examinee information modules, and compares them to verify whether the current examinee is the examinee himself / herself. After the examinee confirms that the examinee is the examinee himself / herself, the examination begins, the examination area is displayed on the display touch screen 3, the audio module 8 instructs the examinee on how to take the examination, and gives voice prompts for irregular actions or cheating. If it is offline mode, the facial feature information on the server is downloaded to the local computer for storage and verification.

[0048] After the examinee enters the examination process, the mainboard 7 in the automatic invigilation device recognizes the examinee's body information, position information, and multi-person information in real time. After the examinee starts the test, the examinee's behavior is judged based on the identified information to see if there is cheating. For example, the examinee leaves the test area, there are multiple people in the test area, and whether the examinee's movements meet the standards. When the examinee's examination time ends or the examination is automatically detected, the examinee's test process log is recorded and uploaded to the examinee's score and examination process recording module. By verifying the test results and the examination room monitoring process, the examinee is checked again to see if he cheated, and the test results are displayed to the examinee, including whether the test is qualified, the test results, whether he cheated, etc. After the examinee confirms, the test for the examinee is completed. If there is an objection, the test process can be viewed through the video, verified and the results can be corrected, and recorded. The entire examination process forms an examination video and file record for subsequent inspection.

Claims

1. An automatic proctoring device for sports test, characterized in that: include: The chassis is a vertical box with rollers, used to carry the components of the device; A display touch screen is arranged on the panel of the chassis and connected to the mainboard, and is used to select test items and display test results; The card reader is arranged on the inner side of the panel of the chassis and connected to the main board, and is used to read the ID card information of the examinee; The image acquisition module is arranged on the panel of the chassis and connected to the main board, and is used to capture the facial information of the test takers and to photograph the completion of the movements of the physical test items performed by the test takers; The audio module is arranged inside the chassis and connected to the mainboard, and is used to announce the test scores of the examinees in sports events; The main board is arranged inside the chassis, and is used to receive information inputted by the display touch screen, the card reader, and the image acquisition module, and after identifying the examinee, runs the automatic proctoring software for the physical test, determines whether the actions of the test items performed by the examinee meet the standards, measures and counts the results of the actions that meet the standards, and gives the examinee's test scores; The communication module is arranged inside the chassis and connected to the main board, and is used to upload the test results to the host computer through the network; as well as The power supply is installed inside the chassis to provide working power for each part.

2. The automatic invigilation device for sports test according to claim 1 is characterized in that include: The hard disk video recorder is arranged inside the chassis and connected to the image acquisition module and the main board respectively, and is used to record and save the movements of the physical test items performed by the examinee.

3. The automatic invigilation device for sports test according to claim 2 is characterized in that: The sports test automatic proctoring software includes: Candidate information module, used to store and call the candidate's identity information; A face feature extraction module is used to extract face features from the facial information of the examinee acquired by the image acquisition module; A face verification module, used to compare the face features extracted by the face feature extraction module with the identity information of the examinee read in by the card reader and the examinee information retrieved from the examinee information module, so as to verify the authenticity of the examinee; and The candidate's score and test process recording module is used to record the candidate's physical test item completion process, determine whether the candidate's completion actions of the test items meet the standards, measure and count the scores of the completion actions that meet the standards, and give the test scores of the test items taken by the candidate.

4. The automatic invigilation device for sports test according to claim 3 is characterized in that: The method for judging whether the candidate's movements meet the standards is as follows: calibration points are set on the human body's trunk, limbs, hands and feet respectively; during the test, the image acquisition module is used to track and identify the corresponding calibration points of the human body in real time; by judging the relative relationship and motion relationship of each calibration point, the positional relationship parameters between the candidate's corresponding calibrations during the test are calculated; and by comparing them with the positional relationship parameters between the corresponding calibrations of the human body required for the standard movements of the tested sports item, a judgment result on whether the candidate's movements completed in this test item meet the standards is obtained.

5. The automatic invigilation device for sports test according to claim 4 is characterized in that: The set calibration points include trunk joints, arm joints, leg joints, hand-related points and foot-related points; the calibrated trunk joints include shoulder joints and hip joints; the calibrated arm joints include wrist joints and elbow joints; the calibrated leg joints include knee joints and ankle joints; the calibrated hand-related points include the fingertips, first joints, second joints and palm joints of each finger; the calibrated foot-related points include toes and heels.

6. A method for unmanned automatic proctoring of sports test, characterized in that: The following steps are involved: S1. Setting the automatic proctoring device for sports test as described in any one of claims 1 to 5 to transmit the identity information of the examinee to the examinee information module; S2. Set the automatic proctoring device for sports test in the examination room, adjust the information acquisition area of ​​the image acquisition module to the test area or the sports facilities for test, select the test sports through the display touch screen, and select the test sports and their test standards according to the gender and age group of the examinee; S3, reading the candidate's identity information through the card reader, collecting the candidate's facial information through the image acquisition module, extracting the candidate's facial feature information through the facial feature extraction module, and after identifying the candidate, storing the above information and information collection time in the candidate information module; S4. Use the image acquisition module to record the entire process of the test items completed by the examinee; based on the video information of the test items completed by the examinee, combined with the automatic monitoring software for sports test running on the mainboard, the completion actions of the test items performed by the examinee are judged to be up to standard, the scores of the completion actions that meet the standards are measured and counted, and the test scores of the test items performed by the examinee are given.

7. The unmanned automatic proctoring method for sports test according to claim 5 is characterized in that: The method for judging whether the candidate's movements are up to standard is: calibrate the joints of the human trunk, limbs, hands and feet respectively, and use the image acquisition module to track and identify the calibrated joints of the human body in real time during the test. By judging the relative relationship and movement relationship of each joint, the candidate's joint parameters during the test are calculated and identified. By comparing them with the human joint parameters required for the standard movements of the tested sports, a judgment result on whether the candidate's movements are up to standard is obtained.

8. The unmanned automatic proctoring method for sports test according to claim 5 is characterized in that: The image acquisition module processes the collected test action videos of the candidates by diverting the video information into multiple video information streams, corresponding to work tasks including face recognition, limb recognition, hand recognition, body recognition, multi-target body recognition, and real-time tracking of body posture recognition, and synchronizes and coordinates the video streams through a synchronizer.

Images