An intelligent system and method for individual tactical training and assessment

Abstract

The present invention relates to an intelligent assessment system, and specifically to an intelligent system and method for individual tactical training and assessment. The present invention solves the technical problems of low efficiency, inaccurate results, and non-standard, non-automatic, and non-intelligent assessment modes in existing tactical training assessments. The system of the present invention includes a video image processing unit, a terminal, a background management center, a crawling unit, an obstacle unit, and a sprint unit; the crawling unit includes a motion posture monitoring module, an acquisition control module, a low-frame crawling module, and a high-frame crawling module; the video image processing unit includes one or more of a face recognition module, a head posture scoring module, a limb movement scoring module, a video monitoring unit, and a data acquisition and analysis center; the data collected by the face recognition module, the video monitoring unit, the head posture scoring module, and the limb movement scoring module are all transmitted to the data acquisition and analysis center. The present invention also provides a tactical training and assessment method based on the system of the present invention.

Description

Technical Field

[0001] The present invention relates to an intelligent assessment system, and in particular to an intelligent system and method for individual tactical training and assessment. Background Art

[0002] Individual tactics are the basic actions of soldiers in combat. Existing individual training usually relies on commanders to use stopwatches to record the completion time of test personnel's actions, and observe the degree of standardization of actions to evaluate the performance. In addition, the collection and storage of training data is an important part of the organization and implementation of military training. Since individual tactical training data is usually collected and recorded manually, its collection and storage efficiency is low and the error rate is high, which is not conducive to the subsequent comprehensive analysis and comparison of data. The traditional individual tactical training and assessment intelligent system uses barbed wire to simulate the actual combat environment, and uses whether the trainees touch the barbed wire as the basis for judgment, which is quite different from the actual combat environment.

[0003] In summary, the existing manual recording method for individual soldier training assessment has the following shortcomings:

[0004] 1. The recording and assessment method is greatly influenced by human factors, resulting in inaccurate assessment results and low efficiency;

[0005] 2. There is a lack of standardized, automated and intelligent objective recording and assessment systems and methods. Summary of the invention

[0006] The purpose of the present invention is to solve the technical problems in the existing individual tactical training and assessment process, such as the traditional manual recording and assessment method leading to inaccurate assessment results and low efficiency, the large difference between the traditional barbed wire and the actual combat environment, and the lack of a standardized, automated and intelligent assessment system and assessment method for the assessment mode, and to provide an individual tactical training and assessment intelligent system and method.

[0007] The technical solution of the present invention:

[0008] An intelligent system for individual tactical training and assessment includes a video image processing unit, a terminal, a background management center, and a crawling unit, an obstacle unit, and a sprint unit which are arranged in sequence;

[0009] Its special features are:

[0010] The terminal includes at least one handheld terminal;

[0011] The crawling unit includes a motion posture monitoring module, a collection control module, and 1 to 4 low-frame crawling modules and 1 to 4 high-frame crawling modules arranged in sequence;

[0012] The motion posture monitoring module includes a plurality of timing sensor groups and a plurality of foul detection sensor groups; the plurality of timing sensor groups and the plurality of foul detection sensor groups are respectively arranged on the frame of the corresponding crawling module, the plurality of timing sensor groups are used to record the time when the trainees enter and exit the net, and the plurality of foul detection sensor groups are used to record the number of times the trainees touch the net illegally;

[0013] The acquisition control module is used to collect detection signals of multiple timing sensor groups and multiple foul detection sensor groups and send them to the handheld terminal;

[0014] The video image processing unit includes one or more of a face recognition module, a head posture scoring module, a body movement scoring module, a video monitoring unit, and a data acquisition and analysis center; the data collected by the face recognition module, the video monitoring unit, the head posture scoring module, and the body movement scoring module are all transmitted to the data acquisition and analysis center;

[0015] The data collection and analysis center is used to control the collection method of the face recognition module, the head posture scoring module and the body movement scoring module, and to pre-process, analyze, count and record the received face recognition data, video surveillance data, head posture scoring data and body movement scoring data, and transmit the data analysis and processing results to the background management center;

[0016] The face recognition module includes a face information collection camera for collecting face information, face recognition and automatic matching of student information;

[0017] The head posture scoring module is used to obtain the student's head image by image recognition in the global coordinate system, calculate the head azimuth, and determine whether the yaw angle, roll angle, and pitch angle of the head at the current moment fall within the corresponding set range;

[0018] The limb action scoring module uses human body morphology to perform human body action recognition and compares human body actions with standard actions; the human body morphology includes appearance, depth, optical flow and human skeleton;

[0019] The obstacle unit is used to assess the enemy situation observation capability;

[0020] The sprint unit is used to detect whether the examinee has reached the finish line and record the sprint time;

[0021] The background management center is used to send the information of the examinees to the handheld terminal via the wireless network, receive the examinees' examination results transmitted by the handheld terminal, receive the video processing results transmitted by the video image processing unit, analyze and count the final examination results, and send the final examination results to the on-site real-time display terminal.

[0022] Further, the video monitoring unit includes a low / high posture crawling monitoring camera, an obstacle monitoring camera and a sprint monitoring camera;

[0023] The low / high posture crawling monitoring camera is used to obtain the original image video of the trainee's crawling posture during the period from low posture entry to high posture exit;

[0024] The obstacle monitoring camera is used to obtain original image videos of trainees' tactical actions using terrain and objects;

[0025] The sprint monitoring camera is used to obtain original image videos of the trainees’ tactical actions in the sprint area;

[0026] The video monitoring unit transmits the acquired original image video to the data acquisition and analysis center, and after being analyzed and processed by the algorithm of the data acquisition and analysis center, it is transmitted to the background management center;

[0027] The terminal also includes at least one on-site real-time display terminal, which is used to display the assessment results;

[0028] The motion posture monitoring module also includes at least two module cameras for collecting the head image of the trainee.

[0029] Further, it also includes NVR;

[0030] The NVR is connected to the data acquisition and analysis center and the background management center respectively. The data acquisition and analysis center marks the original image and video clips, and the marked original image and video clips are transmitted to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image and video clips through the marked timestamp.

[0031] Furthermore, the head posture scoring module implements head posture scoring by the following steps:

[0032] 1) Constructing a first backbone network model to collect and recognize the head image of the trainee in real time; the first backbone network model is a Darknet-53 backbone network;

[0033] 1.1) The Darknet-53 backbone network includes a plurality of convolution blocks and fully connected layers connected in sequence from the input end to the output end; each convolution block includes a convolution layer connected in sequence, a batch normalization layer and an activation layer set in parallel; the convolution layer is composed of 1x1 and 3x3 convolution kernels; when the convolution step size is 1 by default, the padding is consistent by default; when the convolution step size is 2, the padding is valid;

[0034] 1.2) Downsample the Darknet-53 backbone network with a sampling step of 1 or 2 each time; then concatenate the upsampled middle layer and the penultimate layer of the Darknet-53 backbone network to form a new optimized network, which is used to continuously iterate new sample training and learning to obtain the head model;

[0035] 1.3) Optimizing and training the head model through a loss function; the loss function is one or more of target positioning offset loss, target confidence loss, and target classification loss;

[0036] 1.4) Collect the head image information of the trainee to be tested, transmit it to the head model, and recognize the trainee's head image;

[0037] 2) Calculate the head orientation angle

[0038] 2.1) constructing a second backbone network model, wherein the second backbone network model is a ResNet50 backbone network; the ResNet50 backbone network includes three fully connected layers, and the three fully connected layers correspond to the yaw angle, roll angle and pitch angle of the head azimuth respectively; the mapping interval range of the ResNet50 backbone network is [-99, +99]; each fully connected layer makes a separate prediction within the mapping interval, and the fully connected numbers for separate predictions are all binary numbers;

[0039] 2.2) Train the constructed ResNet50 backbone network to obtain the trained ResNet50 backbone network and obtain the actual labeled data;

[0040] 2.3) The head image of the student identified in 1.4) is transmitted to the ResNet50 backbone network trained in step 2.2) to calculate the head orientation angle;

[0041] 2.3.1) Map the yaw angle, roll angle or pitch angle of the head azimuth in the student's head image to the interval [-99, +99] respectively, and classify the 199 values ​​in the mapping interval range from -99 to +99 in sequence, three in a group, and the middle interval as a discrete category;

[0042] 2.3.2) The discrete categories obtained in step 2.3.1) are normalized into probability values ​​after being operated by the softmax function, and the probability values ​​of all discrete category data in the fully connected layer are added to 1;

[0043] 2.3.3) Multiply the probability value obtained in step 2.3.2) by the corresponding discrete category and sum them to obtain the expected value; calculate the mean square error between the expected value and the angle of the actual labeled data, and finally output the head orientation angle through end-to-end training;

[0044] 3) Determine whether the yaw angle, roll angle or pitch angle of the head azimuth at the current moment falls within the corresponding set range; if not, perform corresponding deduction operations according to the deduction rules and give a score; if so, give a score directly.

[0045] Furthermore, the body movement scoring module implements body movement scoring through the following steps:

[0046] 1) Collect students’ human posture images in real time and identify target images;

[0047] 1.1) Establish the first backbone network model;

[0048] 1.2) Extracting 8, 16, and 32 times downsampled features from the first backbone network model constructed in step 1.1);

[0049] 1.3) The features extracted in step 1.2) with 8, 16, and 32 times downsampling are input into the feature pyramid network for multi-scale feature fusion;

[0050] 1.4) Outputting the fusion result of step 1.3) to the detection head to detect and identify the target image;

[0051] 2) Evaluate human posture;

[0052] 2.1) Construct a lightweight network; the lightweight network consists of six layers of convolution:

[0053] The first layer is the Conv / Relu layer, with a core size of 3×3, a stride of 2, and a channel of 32;

[0054] The second layer is the Conv / Relu layer, with a core size of 3×3, a stride of 1, and 64 channels;

[0055] The third layer is the Conv / Relu layer, with a core size of 3×3, a stride of 2, and 64 channels;

[0056] The fourth layer is the Conv / Relu layer, with a core size of 3×3, a stride of 1, and 64 channels;

[0057] The fifth layer is the TransConv / BN / Relu layer, with a core size of 2×2, a stride of 2, and 64 channels;

[0058] The sixth layer is the TransConv / BN / Relu layer, with a core size of 2×2, a stride of 2, and a channel of 32;

[0059] 2.2) Transferring the potential posture from a pre-trained teacher model to a constructed lightweight network to form a student model through a fast posture distillation model training method; the fast posture distillation model training method is to discriminate through a heat map of key bone nodes of the human body;

[0060] 2.3) Downsample the target image detected and identified in step 1.4) twice through strided convolution;

[0061] 2.4) The potential posture learned in the teacher model is used to train the student model, and the target image obtained by two downsamplings in step 2.3) is transmitted to the student model to obtain the confidence of the target image; then the confidence of the target image is compared with the preset confidence standard value. If it is greater than the confidence standard value, an image display of the heat map of the key bone nodes of the human body is output, and then the scoring system is entered and the score is output; if it is less than the confidence standard value, a corresponding point deduction operation is performed, and then the scoring system is entered and the score is output.

[0062] Furthermore, the frame structure of the low frame creeping module is consistent with that of the high frame creeping module;

[0063] The low frame crawling module includes a reinforcing rib assembly and legs, a connecting beam and a horizontal connecting seat arranged on the left and right sides;

[0064] The supporting legs include front legs, middle legs and rear legs; the connecting beam includes front beam and rear beam; the horizontal connecting seat includes front seat, middle seat and rear seat; wherein the top of the front leg is fixedly connected to the front end of the front seat, the top of the middle leg is fixedly connected to the middle of the middle seat, and the top of the rear leg is fixedly connected to the rear end of the rear seat; the rear end of the front seat and the front end of the front beam, the rear end of the front beam and the front end of the middle seat, the rear end of the middle seat and the front end of the rear beam, and the rear end of the rear beam and the front end of the rear seat are respectively hinged through the connecting beam hinge shaft and fixed by positioning pins;

[0065] The reinforcing rib assembly includes three groups of connecting rods respectively arranged between the two front seats on the left and right sides, between the two middle seats and between the two rear seats; each group of connecting rods includes a hinged left rod and a right rod, a left sleeve and a middle sleeve sleeved on both ends of the left rod, and a right sleeve sleeved on the right end of the right rod; the left end of the left rod is hinged to the horizontal connecting seat on the left; the right end of the right rod is hinged to the horizontal connecting seat on the right; the left sleeve is used to fix the hinge between the left rod and the left horizontal connecting seat; the middle sleeve is used to fix the hinge between the left rod and the right rod; the right sleeve is used to fix the hinge between the right rod and the right horizontal connecting seat; wherein, the three hinge axes of the connecting rod between the two front seats are parallel to the front seats; the three connecting beam hinge axes of the connecting rod between the two middle seats are parallel to the middle seats; the three connecting beam hinge axes of the connecting rod between the two rear seats are parallel to the rear seats.

[0066] The number of the timing sensor groups is three; the three timing sensor groups are respectively arranged on the inner sides of the two legs at the front end, the two legs at the change of posture and the two legs at the rear end, and are used for timing the trainer's entry, change of posture and exit from the net; the multiple foul detection sensor groups are respectively arranged on the inner sides of the left and right connecting beams of each low-frame crawling module and the high-frame crawling module, and are used for foul action detection; the module cameras are respectively arranged in the middle of a group of connecting rods at different positions, and are used to collect the trainer's head image;

[0067] The acquisition control module includes at least one acquisition control box and a bracket aviation socket arranged on the outer side of the connecting beam of each low-frame crawling module and the high-frame crawling module; the acquisition control box is electrically connected to the bracket aviation socket through a cable; the bracket aviation socket is electrically connected to the timing sensor group and the foul detection sensor group through a cable.

[0068] Further, the sprint unit includes two sprint rods, a collection control box and at least one photoelectric sensor;

[0069] Each sprint rod is provided with a sprint rod navigation socket or one of the sprint rods is provided with two sprint rod navigation sockets;

[0070] The photoelectric sensor is arranged on the sprint pole to detect the signal of the trainee passing the finish line;

[0071] The acquisition control box is electrically connected to the sprint pole navigation socket via a cable;

[0072] The sprint rod navigation socket is electrically connected to the photoelectric sensor through a cable.

[0073] Furthermore, the comparison of the human body motion with the standard motion is achieved by calculating the Euclidean distance between the human body motion and the standard motion;

[0074] The timing sensor group and the foul detection sensor group are both laser beam type photoelectric switch pairs;

[0075] The two sprint poles are two hollow sprint poles, each of which is higher than two meters; the upper ends of the two hollow sprint poles are connected by a hollow fixed balance pole;

[0076] The photoelectric sensor is a pair of laser beam photoelectric switches, and the receiving end and the transmitting end of the laser beam photoelectric switch are respectively mounted on two hollow sprint rods and are aligned toward each other;

[0077] The crawling unit also includes a buzzer arranged in the acquisition control box;

[0078] The handheld terminals include combat sports handheld terminals and terrain and object handheld terminals;

[0079] The first and acquisition control boxes are both provided with single-chip microcomputers, and all laser-beam photoelectric switches are connected to the GPIO port of the single-chip microcomputer for signals, and then the data are sent to the portable handheld terminal wirelessly.

[0080] A method for individual tactical training and assessment, based on the aforementioned assessment intelligent system, is special in that it includes the following steps:

[0081] 1) Face recognition

[0082] 1.1) Collect students’ facial information through facial information collection cameras and transmit them to the data collection and analysis center;

[0083] 1.2) The data collection and analysis center will structure the student facial information collected in the early stage through face detection, face alignment, feature extraction, and feature matching, and upload the stored data to the background management center;

[0084] 1.3) Automatically identify and match student information, and once the match is successful, enter the crawling unit assessment;

[0085] 2) Crawling unit assessment

[0086] 2.1) The acquisition control box in the crawling unit enters the power-on initialization and self-test mode;

[0087] 2.2) The acquisition control box is connected to the combat sports handheld terminal via wireless. If the connection is abnormal, the buzzer alarm is triggered and the system returns to the power-on initialization and self-test mode. If the connection is normal, the infrared alignment test of the laser beam photoelectric switch pair is performed. If the test is abnormal, the status is sent to the combat sports handheld terminal and the system returns to the power-on initialization and self-test mode. If the test is normal, the system enters the test mode.

[0088] 2.3) The backend management center sends the information of the students to be tested who have been successfully matched in step 1.3) to the combat sports handheld terminal through the wireless network, and the students enter the waiting state; at the same time, the low / high posture crawling monitoring camera is turned on to obtain the original image video of the students crawling during the period from low posture network entry to high posture network exit;

[0089] 2.4) Entering the crawling unit assessment mode, when the assessment trainee enters the network, the first timing sensor group is triggered, and the acquisition control box connected to it starts timing, and the module camera starts shooting at the same time; the acquisition control box uploads the test signal to the combat sports handheld terminal; when the assessment trainee moves forward to the posture change position, the second timing sensor group is triggered, and the acquisition control box records the posture change time and uploads the recorded time to the combat sports handheld terminal; when the assessment trainee leaves the network, the third timing sensor group is triggered, and the acquisition control box records the time of leaving the network and transmits the recorded time to the combat sports handheld terminal; once the assessment trainee touches the laser beam of the upper foul detection sensor group during the crawling process, a foul behavior is generated, the corresponding indicator light on the acquisition control box lights up, and a buzzer alarm is issued to remind the trainee to adjust the posture, and the foul information is transmitted to the combat sports handheld terminal;

[0090] 2.5) After the assessment students leave the network, the crawling unit assessment is completed, and the combat sports handheld terminal sends the assessment results to the background management center; the images captured by the module camera are transmitted to the data acquisition and analysis center for analysis and processing, and then the data information processed by the data acquisition and analysis center is transmitted to the background management center; the original image video obtained by the low / high posture crawling monitoring camera is sent to the data acquisition and analysis center, and after analysis and processing, it is transmitted to the background management center;

[0091] 3) Obstacle unit assessment

[0092] 3.1) Enter the obstacle unit assessment and turn on the obstacle monitoring camera to obtain the original image video of the trainees' tactical actions using terrain features;

[0093] 3.2) After the obstacle unit assessment is completed, the original image video obtained by the obstacle monitoring camera is transmitted to the data acquisition and analysis center, and then transmitted to the background management center after analysis and processing;

[0094] 4) Sprint unit assessment

[0095] 4.1) The acquisition control box in the sprint unit enters the power-on initialization and self-test mode;

[0096] 4.2) The acquisition control box is connected to the terrain handheld terminal via wireless. If the connection is abnormal, the buzzer alarm will be triggered and the system will return to the power-on initialization and self-test mode. If the connection is normal, the infrared alignment test of the photoelectric sensor will be performed. If the test is abnormal, the status will be sent to the terrain handheld terminal and the system will return to the power-on initialization and self-test mode. If the test is normal, the system will enter the waiting state.

[0097] 4.3) The backend management center sends the information of the students to be tested who have been successfully matched in step 1.3) to the terrain handheld terminal through the wireless network, and the students enter the test waiting state; at the same time, the sprint monitoring camera is turned on to obtain the original image video of the tactical actions of the students in the sprint area;

[0098] 4.4) Enter the sprint unit assessment mode

[0099] When the examinee reaches the finish line and the infrared laser beam is blocked, the acquisition control box automatically records the current end time and analyzes and judges the recorded current end time. If the result is unsatisfactory, a buzzer alarm will sound. At the same time, the recorded time information and alarm information will be transmitted wirelessly to the terrain handheld terminal.

[0100] 4.5) The terrain and object handheld terminal sends the assessment results to the background management center, and transmits the original image video obtained by the sprint monitoring camera to the data acquisition and analysis center, and then transmits it to the background management center after analysis and processing;

[0101] 5) The backend management center analyzes and compiles the crawling unit assessment results, obstacle unit assessment results, and sprint unit assessment results of the examinees, and sends the final assessment results to the on-site real-time display terminal.

[0102] Furthermore, the step 2) also includes step 2.6);

[0103] 2.6) The original image video acquired by the low / high posture crawling surveillance camera is sent to the data acquisition and analysis center, which marks the original image video clips, and transmits the marked original image video clips to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image video clips through the marked timestamp;

[0104] The step 3) also includes step 3.3);

[0105] 3.3) The original image video acquired by the obstacle monitoring camera is transmitted to the data acquisition and analysis center, which marks the original image video clips, and transmits the marked original image video clips to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image video clips through the marked timestamp;

[0106] The step 4) also includes step 4.6);

[0107] 4.6) The original image video obtained by the sprint monitoring camera is transmitted to the data acquisition and analysis center, which marks the original image video clips, and transmits the marked original image video clips to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image video clips through the marked timestamp.

[0108] Beneficial effects of the present invention:

[0109] 1. The present invention adopts modern optoelectronic technology, video AI processing, storage and analysis of massive data, signal processing technology, software technology, Internet of Things and big data technology to improve the autonomy, scientificity, standardization and accuracy of training assessment, save manpower and improve assessment efficiency, and better protect the life safety of trainees.

[0110] 2. The training and assessment of the present invention is realized electronically. From the beginning, process to the end, the detection of the entire training and assessment process adopts photoelectric technology and wireless communication technology, and the data processing is automated to ensure the timeliness and accuracy of the data.

[0111] 3. The training and assessment of the present invention has the advantage of autonomy. With the support of an intelligent training management system, trainees can independently apply for and complete training, and the results are automatically recorded and transmitted, thereby ensuring the fairness of training and assessment.

[0112] 4. The training and assessment process of the present invention is monitored and alarmed in a timely manner, effectively avoiding injuries to trainees and providing early warnings for potential health and safety hazards in training; trainees can carry out targeted intensive training to prevent accidents and effectively protect their lives.

[0113] 5. The present invention uses AI intelligent algorithms to determine whether a single soldier's action is a foul, achieves objective quantification and standardized evaluation criteria, makes the results accurate and fair, and greatly improves work efficiency.

[0114] 6. The high and low creeping bracket of the present invention is foldable, which is convenient for installation, disassembly and transportation.

[0115] 7. The background management software of the present invention can realize data storage, display and analysis, and is simple to operate and highly efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0116] Figure 1 It is a connection relationship diagram of the assessment intelligent system of the present invention (video monitoring unit, including video monitoring a for low / high posture crawling monitoring camera; video monitoring b for obstacle monitoring camera; video monitoring c for sprint monitoring camera);

[0117] Figure 2 is a schematic structural diagram of a crawling unit in an embodiment of the present invention;

[0118] Figure 3 is a structural schematic diagram of a low-frame crawling module of a crawling unit in an embodiment of the present invention;

[0119] Figure 4 yes Figure 3 A partial enlarged view of

[0120] Figure 5 is a schematic diagram of the structure of a sprint unit in an embodiment of the present invention;

[0121] Figure 6 It is a connection relationship diagram of the video monitoring unit, the video image processing unit and the background management center in the embodiment of the present invention (the video monitoring unit includes video monitoring a as a low / high posture crawling monitoring camera; video monitoring b as an obstacle monitoring camera; video monitoring c as a sprint monitoring camera);

[0122] Figure 7 is a connection relationship diagram between a data acquisition and analysis center and a processing and analysis object in an embodiment of the present invention;

[0123] Figure 8 It is a connection relationship diagram of the crawling unit, the sprinting unit and the background management center in the embodiment of the present invention;

[0124] Fig. 9 It is a software design flow chart of the acquisition control box of the crawling unit in the embodiment of the present invention;

[0125] Fig.10 It is a software design flow chart in the acquisition control box of the sprint unit in an embodiment of the present invention;

[0126] Fig.11 is a functional block diagram of a handheld terminal in an embodiment of the present invention;

[0127] Fig.12 This is a flowchart of the handheld terminal software design in an embodiment of the present invention;

[0128] Fig.13 This is a functional block diagram of the background management software in an embodiment of the present invention;

[0129] Fig.14 This is a flow chart of the background management software in an embodiment of the present invention;

[0130] Fig.15 is a schematic diagram of a fully folded structure of a low-frame crawling module of a crawling unit in an embodiment of the present invention;

[0131] Fig.16 : is a schematic diagram of the network structure of the ResNet50 backbone network in the present invention, where MSE is the mean square error;

[0132] Fig.17It is the flow chart of MBConv1 in the present invention;

[0133] Fig.18 It is the flow chart of MBConv6 in the present invention;

[0134] Fig.19 It is a flow chart of human posture assessment in the present invention.

[0135] Description of reference numerals:

[0136] 1- low-frame crawling module, 101- front legs, 102- middle legs, 103- rear legs, 104- front beam, 105- rear beam, 106- front seat, 107- middle seat, 108- rear seat, 109- left rod, 110- right rod, 111- left sleeve, 112- middle sleeve, 113- right sleeve, 114- height adjustment seat, 115- connecting beam hinge shaft, 116- positioning pin; 3- high-frame crawling module; 2- acquisition control box, 201- module camera, 202- timing sensor group, 203- foul detection sensor group, 204- bracket navigation socket; 22- sprint rod, 23- photoelectric sensor, 24- sprint rod navigation socket, 25- hollow fixed balance rod, 26- rectangular sleeve. DETAILED DESCRIPTION

[0137] The present invention provides an intelligent system for individual tactical training and assessment, such as Figure 1 As shown, it includes a video image processing unit, a terminal, a background management center, an NVR, and also includes a crawling unit, an obstacle unit and a sprinting unit arranged in sequence; the terminal includes two handheld terminals and a field real-time display terminal. The two handheld terminals are respectively a combat movement handheld terminal and a terrain and feature handheld terminal. The combat movement handheld terminal is used for the crawling unit, and the terrain and feature handheld terminal is used for the sprinting unit.

[0138] like Figure 2 As shown, the crawling unit includes a motion posture monitoring module, a collection control module, and four low-frame crawling modules 1 and four high-frame crawling modules 3 arranged in sequence from low to high.

[0139] The frame structures of the low frame crawling module 1 and the high frame crawling module 3 are the same, the difference is that the heights of the legs are different. This embodiment only describes the structure of the low frame crawling module 1 in detail. Figure 3 and Figure 4 As shown, the low frame crawling module 1 includes a reinforcing rib assembly and legs, connecting beams and horizontal connecting seats arranged on the left and right sides of the reinforcing rib assembly. The dimensions of the elevated crawling module 3 are: length × width × height = 2500mm × 1800mm × 600mm, and the dimensions of the low frame crawling module 1 are: length × width × height

[0140] =2500mm×1800mm×400mm.

[0141] The supporting legs include a front leg 101, a middle leg 102 and a rear leg 103; the connecting beam includes a front beam 104 and a rear beam 105; and the horizontal connecting seat includes a front seat 106, a middle seat 107 and a rear seat 108. Among them, the top of the front leg 101 is fixedly connected to the front end of the front seat 106, the top of the middle leg 102 is fixedly connected to the middle of the middle seat 107, and the top of the rear leg 103 is fixedly connected to the rear end of the rear seat 108; the rear end of the front seat 106 and the front end of the front beam 104, the rear end of the front beam 104 and the front end of the middle seat 107, the rear end of the middle seat 107 and the front end of the rear beam 105, and the rear end of the rear beam 105 and the front end of the rear seat 108 are respectively hinged through a connecting beam hinge shaft 115 and fixed by a positioning pin 116. The sum of the lengths of the middle leg 102 and the front leg 101 is less than the length of the front beam 104, and the sum of the lengths of the middle leg 102 and the rear leg 103 is less than the length of the rear beam 105. A height adjustment seat 114 is provided at the bottom of the front leg 101 , the middle leg 102 and the rear leg 103 .

[0142] The reinforcing rib assembly includes three groups of connecting rods respectively arranged between the two front seats 106 on the left and right sides, between the two middle seats 107 and between the two rear seats 108, each group of connecting rods includes a hinged left rod 109 and a right rod 110, a left sleeve 111 and a middle sleeve 112 sleeved on both ends of the left rod 109, and a right sleeve 113 sleeved on the right end of the right rod 110. The left end of the left rod 109 is hinged to the horizontal connecting seat on the left, and the right end of the right rod 110 is hinged to the horizontal connecting seat on the right. The left sleeve 111 is used to fix the hinge between the left rod 109 and the left horizontal connecting seat, the middle sleeve 112 is used to fix the hinge between the left rod 109 and the right rod 110, and the right sleeve 113 is used to fix the hinge between the right rod 110 and the right horizontal connecting seat. Among them, the three hinge axes of the connecting rod between the two front seats 106 are parallel to the front seats 106, the three hinge axes of the connecting rod between the two middle seats 107 are parallel to the middle seats 107, and the three hinge axes of the connecting rod between the two rear seats 108 are parallel to the rear seats 108. The length of the left rod 109 is less than the length of the front beam 104 and the length of the rear beam 105; the length of the right rod 110 is less than the length of the front beam 104 and the length of the rear beam 105. The low / high frame crawling module can be folded, and the low frame crawling module after being fully folded is as follows Fig.15 shown.

[0143] The motion posture monitoring unit includes four module cameras 201, three timing sensor groups 202, and multiple foul detection sensor groups 203. The module cameras 201 are respectively arranged in the middle of the rear end connecting rods of the second and fourth low-frame crawling modules 1, and the middle of the rear end connecting rods of the second and fourth high-frame crawling modules 3, and are used to shoot the motion postures of the trainer at each stage in the net, especially the head image of the trainer, for the judgment of the enemy's concept. The module camera 201 is fixed by butterfly screws, and it is easy to disassemble and install. The three timing sensor groups 202 are respectively arranged on the inner side surfaces of the two legs at the front end, the two legs at the change of posture, and the two legs at the rear end, for the timing of the trainer entering the net, changing posture and leaving the net. Multiple foul detection sensor groups 203 are evenly arranged on the inner side surfaces of the left and right connecting beams of each low-frame crawling module 1 and the high-frame crawling module 3, for foul action detection. The timing sensor group 202, the foul detection sensor group 203 and the module camera trigger sensor group are all laser beam type photoelectric switches.

[0144] The acquisition control module includes at least one acquisition control box 2 and a bracket navigation socket 204 arranged on the outer side of the connecting beam of each low-frame crawling module 1 and the high-frame crawling module 3; the acquisition control box 2 is electrically connected to the bracket navigation socket 204 through a cable; the bracket navigation socket 204 is electrically connected to the timing sensor group 202 and the foul detection sensor group 203 through a cable. A single-chip microcomputer is arranged in the acquisition control box 2, and all laser-beaming photoelectric switch signals are connected to the GPIO port of the single-chip microcomputer, and then the data is sent to the combat sports handheld terminal wirelessly. A buzzer is arranged in the acquisition control box 2. When students make illegal actions during crawling training, there will be a buzzer alarm prompt so that students can adjust their training posture in time. The software design process in the acquisition control box of the crawling unit is as follows: Fig. 9 As shown,

[0145] The sprint unit is used to detect whether the examinee has reached the finish line and record the sprint time. Figure 5As shown, the sprint unit includes two sprint rods 22, a collection control box 2, a terrain and object handheld terminal, and two photoelectric sensors 23; in this embodiment, the photoelectric sensor 23 is a laser beam photoelectric switch. The two sprint rods 22 are two hollow sprint rods, and the hollow sprint rods are higher than two meters; the upper ends of the two hollow sprint rods are connected by a hollow fixed balance rod 25; the upper ends of the hollow sprint rods are connected to the hollow fixed balance rod 25 through a rectangular sleeve 26. Two pairs of laser beam photoelectric switches are distributed from top to bottom along the hollow sprint rods; the laser beam photoelectric switch pair consists of a receiving end and a transmitting end, which are respectively installed on the two hollow sprint rods and aligned with each other; the receiving end of the laser beam photoelectric switch is on one hollow sprint rod, and the transmitting end is on the other hollow sprint rod. Two sprint rod navigation sockets 24 are provided on one of the hollow sprint rods. In other embodiments, one sprint rod navigation socket 24 may be provided on each hollow sprint rod. The acquisition control box 2 is electrically connected to the sprint rod navigation socket 24 via a cable. The sprint rod navigation socket 24 is electrically connected to the laser-beaming photoelectric switch via a cable. A single-chip microcomputer is provided in the acquisition control box 2. All signals of the laser-beaming photoelectric switch are connected to the GPIO port of the single-chip microcomputer, and then the data is sent to the terrain handheld terminal wirelessly. A buzzer is provided in the acquisition control box 2, and a buzzer warning is issued if the trainee fails the sprint time. The software design process in the acquisition control box 2 of the sprint unit is as follows. Fig.10 shown.

[0146] The combat sports handheld terminal and the terrain handheld terminal are respectively associated with the background management center by wireless means, and can upload and download assessment information, query the score ranking, etc. The handheld terminal includes a shell, an MCU module, a wireless communication module, a buzzer and a rechargeable battery arranged in the shell, an LCD touch screen arranged on the front of the shell, a capacitive pen unit, a power switch and a charging port arranged on the bottom surface of the shell, a hand strap and an antenna arranged on the top surface of the shell, a hand strap unit arranged on the back of the shell, an anti-slip structure arranged on the left and right sides of the shell, and a buffer pad arranged at the eight corners of the shell; the antenna is electrically connected to the MCU module through the wireless communication module; the charging port is electrically connected to the rechargeable battery; the rechargeable battery is electrically connected to the LCD touch screen and the MCU module through the power switch; the MCU module is electrically connected to the LCD touch screen and the buzzer respectively; the hand strap unit includes a pocket cloth located in the middle of the back of the shell and an elastic band connected to the four corners of the pocket cloth respectively; the other end of the elastic band is fixed to the four corners of the back of the shell respectively.

[0147] The overall interface design of the handheld terminal is optimized, easy to operate, has good human-computer interaction, and its structure conforms to the ergonomic layout. Both portable handheld terminals can transmit information wirelessly with the high and low net crawling unit, the sprint unit, and the background management center. On the one hand, it receives the information detected by the high and low net crawling unit and the sprint unit on-site photoelectric switches (number of net-touching fouls, net entry and exit, and assessment start and end time), and on the other hand, it receives the assessment student information sent by the background management center and transmits the student's on-site scoring results back to the assessment center. It is an important part of the entire training assessment system. The functional diagram of the handheld terminal is shown in the figure below. Fig.11 As shown in the figure, the handheld terminal software design flow chart is as follows Fig.12 shown.

[0148] The obstacle unit is used to assess the trainees' enemy observation ability. It sets up earth piles or pits to examine the trainees' postures such as bending forward, leaping, rolling forward, rolling sideways, observing the enemy, and rushing out. The obstacle monitoring camera is used to obtain the original image video of the trainees' tactical actions using terrain and objects, and transmits it to the limb movement scoring module in the data acquisition and analysis center for preprocessing, analysis and statistics, and record storage. Finally, the data analysis and processing results are transmitted to the background management center.

[0149] The video image processing unit includes one or more of a face recognition module, a head posture scoring module, a data acquisition and analysis center, a limb movement scoring module and a video monitoring unit; the data collected by the face recognition module, the video monitoring unit, the head posture scoring module and the limb movement scoring module are all transmitted to the data acquisition and analysis center. The data acquisition and analysis center includes data acquisition, equipment management, statistical analysis, and resource allocation. It performs face recognition on trainees and manages the face database. It analyzes and quantitatively scores the head posture, bending forward, leaping forward, forward rolling, side rolling, and grabbing postures in the assessment. Finally, it stores various data generated in the assessment and provides data support for the background management center. The connection relationship between the video monitoring unit, the video image processing unit and the background management center is as follows: Figure 6 shown.

[0150] like Figure 7As shown, the data collection and analysis center is used to control the collection method of the face recognition module, the head posture scoring module and the body movement scoring module, and to pre-process, analyze, count and record the received face recognition module data, the video monitoring unit data, the head posture scoring module data and the body movement scoring module data, and transmit the data analysis and processing results to the background management center. In this embodiment, an NVR (network hard disk recorder) is also provided, and the NVR is used to record and store the original video image fragments. Specifically, the data collection and analysis center marks the original image video fragments, and the marked original image video fragments are transmitted to the NVR for recording and storage; at the same time, the data collection and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image video fragments through the marked timestamp.

[0151] The data acquisition and analysis center is the core of data acquisition and data processing, including image acquisition and conditional trigger modules, which can control the acquisition method and analyze the collected data to form an evaluation system for student assessment. Its functions include: corresponding data acquisition trigger conditions, preprocessing of received data, exclusion of invalid data, analysis and statistics of data according to modules, recording of data at each scoring point, data transfer services, and providing them to the background management center, unified coordination and management of each recognition and scoring module, so that it can maximize the use of computing resources. The system adopts the TCP / IP protocol and UDP protocol in a coordinated manner, and builds an application system on it, which not only guarantees the transmission rate but also the transmission quality; it tracks the health status in the front-end acquisition module.

[0152] The face recognition module includes a face information collection camera, which is used to collect face information, face recognition and automatically match student information. Face recognition is divided into four processes: face detection, face alignment, feature extraction, and feature matching. Among them, feature extraction is the most critical step in face recognition. The extracted features are more inclined to the "unique" features of the face, which plays a pivotal role in feature matching. The neural network and model bear the heavy responsibility of extracting features. Excellent network and training strategies make the model more robust. After the student information is collected in the early stage, the facial data is structured and stored through the above process. In the subsequent training and assessment process, the front-end acquisition equipment will automatically match the student information; during the training and assessment, the student only needs to receive the acquisition and recognition ready signal in the assessment preparation area, which saves the time of the assessor to enter the information data and avoids the risk of human input errors as much as possible.

[0153] The head posture scoring module is used to obtain the student's head image through image recognition in the global coordinate system, calculate the head azimuth, and determine whether the yaw angle, roll angle, and pitch angle of the head at the current moment fall within the corresponding set range.

[0154] When students are crawling, they observe whether the front meets the standards based on this technology. We collect the student's head posture through this technology, analyze the angle indicated by the posture in horizontal and vertical sections, calculate the head azimuth, and then determine whether the yaw angle, roll angle or pitch angle of the head at the current moment falls within the corresponding set range, so that the student assessment standards can be quantified.

[0155] The head posture scoring module implements head posture scoring through the following steps:

[0156] 1) Constructing a first backbone network model to collect and recognize the head image of the trainee in real time; the first backbone network model is a Darknet-53 backbone network;

[0157] 1.1) The Darknet-53 backbone network includes a plurality of convolution blocks and fully connected layers connected in sequence from the input end to the output end; each convolution block includes a convolution layer connected in sequence, a batch normalization layer and an activation layer set in parallel; the convolution layer is composed of 1x1 and 3x3 convolution kernels; when the convolution step size is 1 by default, the padding is consistent by default; when the convolution step size is 2, the padding is valid;

[0158] 1.2) Downsample the Darknet-53 backbone network with a sampling step of 1 or 2 each time; then concatenate the upsampled middle layer and the penultimate layer of the Darknet-53 backbone network to form a new optimized network, which is used to continuously iterate new sample training and learning to obtain the head model;

[0159] 1.3) Optimizing and training the head model through a loss function; the loss function is one or more of target positioning offset loss, target confidence loss, and target classification loss;

[0160] 1.4) Collect the head image information of the trainee to be tested, transmit it to the head model, and recognize the trainee's head image;

[0161] 2) Calculate the head orientation angle

[0162] 2.1) Construct the second backbone network model, which is the ResNet50 backbone network; Fig.16 As shown, the ResNet50 backbone network includes three fully connected layers, and the three fully connected layers correspond to the yaw angle, roll angle and pitch angle of the head azimuth respectively; the mapping interval range of the ResNet50 backbone network is [-99, +99]; each fully connected layer makes a separate prediction within the mapping interval, and the fully connected numbers for separate predictions are all binary numbers;

[0163] 2.2) Train the constructed ResNet50 backbone network to obtain the trained ResNet50 backbone network and obtain the actual labeled data;

[0164] 2.3) The head image of the student identified in step 1.4) is transmitted to the ResNet50 backbone network trained in step 2.2) to calculate the head orientation angle;

[0165] 2.3.1) Map the yaw angle, roll angle or pitch angle of the head azimuth in the student's head image to the interval [-99, +99] respectively, and classify the 199 values ​​in the mapping interval range from -99 to +99 in sequence, three in a group, and the middle interval as a discrete category;

[0166] 2.3.2) The discrete categories obtained in step 2.3.1) are normalized into probability values ​​after being operated by the softmax function, and the probability values ​​of all discrete category data in the fully connected layer are added to 1;

[0167] 2.3.3) Multiply the probability value obtained in step 2.3.2) by the corresponding discrete category and sum them to obtain the expected value; calculate the mean square error between the expected value and the angle of the actual labeled data, and finally output the head orientation angle through end-to-end training;

[0168] 3) Determine whether the yaw angle, roll angle or pitch angle of the head azimuth at the current moment falls within the corresponding set range; if not, perform corresponding deduction operations according to the deduction rules and give a score; if so, give a score directly.

[0169] The limb action scoring module uses human body morphology to identify human actions and calculates the Euclidean distance between human actions and standard actions to compare human actions with standard actions; human body morphology such as appearance, depth, optical flow, human skeleton, and dynamic bone mode can be naturally represented by the time series of human joint positions. Then, human behavior recognition can be achieved by analyzing its action pattern. Therefore, the bottom-up human posture estimation of some affinity areas can be used to analyze its actions to obtain the score between its actions and standard actions. For example, in the process of forward roll and side roll, the key points of the human body can be analyzed by a high dynamic camera, so that the spatial domain model can be applied in the time domain, and the computer can identify whether the matching student's action is forward roll or side roll, and calculate the Euclidean distance between the standard action and the collected data to obtain the corresponding score.

[0170] The body movement scoring module performs processing and analysis through the following steps:

[0171] 1) Collect students’ human posture images in real time and identify target images;

[0172] 1.1) Establish the first backbone network model;

[0173] The first backbone network model is composed of MBConv1 and MBConv6. Fig.17 As shown in , MBConv1 includes 1×1 convolution, batch normalization layer and activation layer set in parallel, DWConv k×k, batch normalization layer, activation layer and compression and excitation module set in parallel, 1×1 convolution, batch normalization layer and Add (fusion module), and side connection from input to Add (fusion module); wherein the step size of DWConv k×k is 1; as shown in Fig.18 As shown, MBConv6 includes 1×1 convolution, batch normalization layer and activation layer set in parallel, DWConv k×k, batch normalization layer set in parallel, activation layer and compression and excitation module, 1×1 convolution and batch normalization layer connected in sequence from the input end to the output end; among them, the step size of DWConv k×k is 2.

[0174] It is also possible to optimize the first backbone network model to obtain a more simplified first backbone network model, that is, remove all the original convolutional layers in the first backbone network model, and only retain the 1×1 convolution extracted from the first backbone network model to align the feature dimension and channel. Upsampling and downsampling are completed in an interpolation manner. Finally, the simplified first backbone network model is as follows:

[0175] The operator of the first stage is 3×3 convolution, the resolution is 224×224, the channels are 32, and the number of layers is 1;

[0176] The operator in the second stage is MBConv1, with k3×3, resolution of 112×112, channels of 16, and layers of 1;

[0177] The operator in the third stage is MBConv6, with k3×3, resolution of 112×112, channels of 24, and layers of 2;

[0178] The operator in the fourth stage is MBConv6, with k5×5, resolution 56×56, 40 channels, and 2 layers;

[0179] The operator in the fifth stage is MBConv6, with k3×3, resolution of 28×28, channels of 80, and layers of 3. The simplified backbone network model can greatly reduce the amount of calculation and make the output results fast and accurate.

[0180] 1.2) Extracting 8, 16, and 32 times downsampled features from the first backbone network model constructed in step 1.1);

[0181] 1.3) The features extracted in step 1.2) with 8, 16, and 32 times downsampling are input into the feature pyramid network for multi-scale feature fusion;

[0182] 1.4) Outputting the fusion result of step 1.3) to the detection head to detect and identify the target image;

[0183] The Focal loss function used by the detection head is a distributed Focal loss function.

[0184] 2) Evaluate human posture;

[0185] 2.1) Construct a lightweight network; the lightweight network consists of six layers of convolution:

[0186] The first layer is the Conv / Relu layer, with a core size of 3×3, a stride of 2, and a channel of 32;

[0187] The second layer is the Conv / Relu layer, with a core size of 3×3, a stride of 1, and 64 channels;

[0188] The third layer is the Conv / Relu layer, with a core size of 3×3, a stride of 2, and 64 channels;

[0189] The fourth layer is the Conv / Relu layer, with a core size of 3×3, a stride of 1, and 64 channels;

[0190] The fifth layer is the TransConv / BN / Relu layer, with a core size of 2×2, a stride of 2, and 64 channels;

[0191] The sixth layer is the TransConv / BN / Relu layer, with a core size of 2×2, a stride of 2, and a channel of 32;

[0192] 2.2) Transferring the potential posture from a pre-trained teacher model to a constructed lightweight network to form a student model through a fast posture distillation model training method; the fast posture distillation model training method is to discriminate through a heat map of key bone nodes of the human body;

[0193] 2.3) Downsample the target image detected and identified in step 1.4) twice through strided convolution;

[0194] 2.4) Use the potential posture learned in the teacher model to train the student model, and then press Fig.19The process is as follows: first, the target image obtained by two downsamplings in step 2.3) is transmitted to the student model to obtain the confidence of the target image; then, the confidence of the target image is compared with the preset confidence standard value of 0.85±0.02. If it is greater than the confidence standard value, the image display of the heat map of the key bone nodes of the human body is output, and then the scoring system is entered and the score is output; if it is less than the confidence standard value, the corresponding point deduction operation is performed, and then the scoring system is entered and the score is output.

[0195] The video monitoring unit includes low / high posture crawling monitoring cameras, obstacle monitoring cameras and sprint monitoring cameras; the low / high posture crawling monitoring cameras are used to obtain the original image video of the trainee's crawling posture from low posture entry to high posture exit, and the camera is installed on the steel net of the crawling unit; the obstacle monitoring camera is used to obtain the original image video of the trainee's tactical actions using terrain and objects; the sprint monitoring camera is used to obtain the original image video of the trainee's tactical actions in the sprint area. The camera has strong light suppression, high dynamic, three-proof, and anti-fouling effects, and can achieve 360-degree no-dead-angle monitoring according to the shooting angle requirements.

[0196] The video image processing unit also includes a deployment and adjustment unit. The deployment and adjustment unit is a set of fast debugging units designed for the premise that there are many video image acquisition terminals on site and the debugging is relatively complicated. The main function of this unit is to quickly deploy each time at the assessment site, so it provides an integrated unit with functions such as video accuracy, positioning, focusing, and communication testing. The unit has light storage and three-proof properties to facilitate deployment and storage of each data acquisition and analysis center.

[0197] like Figure 1 and Figure 8 As shown, the backend management center is used to send the information of the assessment students to the handheld terminal through the wireless network, receive the assessment scores of the assessment students transmitted by the handheld terminal, receive the video processing results transmitted by the video image processing unit, analyze and count the final assessment scores, and send the final assessment scores to the on-site real-time display terminal. The backend management center plans the entire assessment system and sends the information of the assessment students to the combat movement handheld terminal and the terrain handheld terminal through the wireless network. After the handheld terminal obtains the assessment scores of the assessment students, it sends them to the backend management center through the wireless network; at the same time, the video monitoring unit transmits the video image to the data acquisition and analysis center through the wired network for sample comparison and analysis. After the video image processing is completed, the video image processing results are sent to the backend management center through a wired method; the backend management center analyzes and counts the final assessment scores, sends the statistically completed assessment scores to the on-site real-time display terminal through the wireless network, and stores the assessment scores in the database.

[0198] Backend management software functions such as Fig.13As shown in the figure, the backend management center mainly includes the assessment software. The assessment software of the backend management center is mainly used to manage the assessment students, assessment content, assessment process, and assessment results, and to record, analyze, and store the assessment and training results of students for daily assessment and training. It also plans, analyzes, and counts the entire assessment and training activities, obtains the final assessment results, and generates the assessment status change table and personal training log of the students, which dynamically reflects the assessment and training status of the students. The backend management software process is as follows Fig.14 shown.

[0199] The on-site real-time display terminal is used to display the assessment results. During the examination, the background management center sends the student's assessment results and assessment status to the on-site real-time display terminal by wireless means for real-time display. In this embodiment, the assessment results are displayed in a rolling table, and the progress, assessment status, and performance evaluation of the assessors are statistically displayed using bar charts, pie charts, line charts, etc., to achieve real-time visualization of the assessment. According to user needs, the single-soldier tactical training and assessment intelligent system mainly uses multiple pairs of photoelectric sensors instead of traditional barbed wire to collect, process, and display in real time the trainer's entry, exit, leap, sprint, and illegal touch signals during the training process, and combines video image recognition technology to compare tactical actions to determine the time it takes for the trainee to pass through all the bunkers and whether the action essentials are standardized. The assessment results are transmitted to the background management center by wireless transmission for analysis and statistics, which is used to improve the timeliness, accuracy, and fairness of single-soldier assessment training.

[0200] The present invention provides a single-soldier tactical training and assessment method, based on the above-mentioned assessment intelligent system, comprising the following steps:

[0201] 1) Face recognition

[0202] 1.1) Collect students’ facial information through facial information collection cameras and transmit them to the data collection and analysis center;

[0203] 1.2) The data collection and analysis center will structure the student facial information collected in the early stage through face detection, face alignment, feature extraction, and feature matching, and upload the stored data to the backend management center;

[0204] 1.3) Automatically identify and match student information, and once the match is successful, enter the crawling unit assessment;

[0205] 2) Crawling unit assessment

[0206] 2.1) If Fig. 9 As shown, the acquisition control box 2 in the crawling unit enters the power-on initialization and self-test mode;

[0207] 2.2) The acquisition control box 2 is connected to the combat sports handheld terminal via wireless. If the connection is abnormal, a buzzer alarm is triggered and the system returns to the power-on initialization and self-test mode. If the connection is normal, the infrared alignment test of the laser beam photoelectric switch pair is performed. If the test is abnormal, the status is sent to the combat sports handheld terminal and the system returns to the power-on initialization and self-test mode. If the test is normal, the system enters the test mode.

[0208] 2.3) The backend management center sends the information of the students to be tested who have been successfully matched in step 1.3) to the combat sports handheld terminal through the wireless network, and the students enter the waiting state; at the same time, the low / high posture crawling monitoring camera is turned on to obtain the original image video of the students crawling during the period from low posture network entry to high posture network exit;

[0209] 2.4) Entering the crawling unit assessment mode, when the assessment trainee enters the network, the first timing sensor group 202 is triggered, and the acquisition control box 2 connected thereto starts timing, and the module camera 201 starts shooting at the same time; the acquisition control box 2 uploads the test signal to the combat sports handheld terminal; when the assessment trainee advances to the position where the posture is changed, the second timing sensor group 202 is triggered, and the acquisition control box 2 records the time of the posture change, and uploads the recorded time to the combat sports handheld terminal; when the assessment trainee leaves the network, the third timing sensor group 202 is triggered, and the acquisition control box 2 records the time of leaving the network, and transmits the recorded time to the combat sports handheld terminal; once the assessment trainee touches the laser beam of the upper foul detection sensor group 203 during the crawling process, a foul behavior is generated, the corresponding indicator light on the acquisition control box 2 lights up, and a buzzer alarm is issued to remind the trainee to adjust the posture, and the foul information is transmitted to the combat sports handheld terminal at the same time;

[0210] 2.5) After the crawling unit assessment is completed, the combat sports handheld terminal sends the assessment results to the background management center; the images captured by the module camera 201 are transmitted to the data acquisition and analysis center for analysis and processing, and then the data information processed by the data acquisition and analysis center is transmitted to the background management center; the original image video obtained by the low / high posture crawling monitoring camera is sent to the data acquisition and analysis center, and transmitted to the background management center after analysis and processing;

[0211] 2.6) The original image video acquired by the low / high posture crawling surveillance camera is sent to the data acquisition and analysis center, which marks the original image video clips, and transmits the marked original image video clips to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image video clips through the marked timestamp;

[0212] 3) Obstacle unit assessment

[0213] 3.1) Enter the obstacle unit assessment and turn on the obstacle monitoring camera to obtain the original image video of the trainees' tactical actions using terrain features;

[0214] 3.2) After the obstacle unit assessment is completed, the original image video obtained by the obstacle monitoring camera is transmitted to the data acquisition and analysis center, and then transmitted to the background management center after analysis and processing;

[0215] 3.3) The original image video acquired by the obstacle monitoring camera is transmitted to the data acquisition and analysis center, which marks the original image video clips, and transmits the marked original image video clips to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image video clips through the marked timestamp;

[0216] 4) Sprint unit assessment

[0217] 4.1) Fig.10 As shown, the acquisition control box 2 in the sprint unit enters the power-on initialization and self-test mode;

[0218] 4.2) The acquisition control box 2 is connected to the terrain handheld terminal via wireless. If the connection is abnormal, a buzzer alarm is triggered and the system returns to the power-on initialization and self-test mode. If the connection is normal, the infrared alignment test of the photoelectric sensor is performed. If the test is abnormal, the status is sent to the terrain handheld terminal and the system returns to the power-on initialization and self-test mode. If the test is normal, the system enters the waiting state.

[0219] 4.3) The backend management center sends the information of the students to be tested who have been successfully matched in step 1.3) to the terrain handheld terminal through the wireless network, and the students enter the test waiting state; at the same time, the sprint monitoring camera is turned on to obtain the original image video of the tactical actions of the students in the sprint area;

[0220] 4.4) Enter the sprint unit assessment mode

[0221] When the student reaches the finish line, the infrared laser beam of the laser photoelectric switch is blocked, and the acquisition control box 2 automatically records the current end time and analyzes and judges the recorded current end time. If the result is unqualified, a buzzer alarm will sound; at the same time, the recorded time information and alarm information will be transmitted wirelessly to the terrain handheld terminal;

[0222] 4.5) The terrain and object handheld terminal sends the assessment results to the background management center, and transmits the original image video obtained by the sprint monitoring camera to the data acquisition and analysis center, and then transmits it to the background management center after analysis and processing;

[0223] 4.6) The original image video obtained by the sprint monitoring camera is transmitted to the data acquisition and analysis center, which marks the original image video clips, and transmits the marked original image video clips to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image video clips through the marked timestamp.

[0224] 5) The backend management center analyzes and compiles the crawling unit assessment results, obstacle unit assessment results, and sprint unit assessment results of the examinees, and sends the final assessment results to the on-site real-time display terminal.

Claims

1. An intelligent system for individual tactical training and assessment, comprising a video image processing unit, a terminal, a background management center, and a crawling unit, an obstacle unit and a sprint unit arranged in sequence; Features: The terminal includes at least one handheld terminal; The crawling unit comprises a motion posture monitoring module, a collection control module, and 1 to 4 low-frame crawling modules (1) and 1 to 4 high-frame crawling modules (3) arranged in sequence; The motion posture monitoring module comprises a plurality of timing sensor groups (202) and a plurality of foul detection sensor groups (203); the plurality of timing sensor groups (202) and the plurality of foul detection sensor groups (203) are respectively arranged on the frame of the corresponding crawling module, the plurality of timing sensor groups (202) are used to record the time when the trainee enters and exits the net, and the plurality of foul detection sensor groups (203) are used to record the number of times the trainee touches the net illegally; The acquisition control module is used to acquire detection signals of a plurality of timing sensor groups (202) and a plurality of foul detection sensor groups (203) and send the detection signals to a handheld terminal; The video image processing unit includes one or more of a face recognition module, a head posture scoring module, a body movement scoring module, a video monitoring unit, and a data acquisition and analysis center; the data collected by the face recognition module, the video monitoring unit, the head posture scoring module, and the body movement scoring module are all transmitted to the data acquisition and analysis center; The data collection and analysis center is used to control the collection method of the face recognition module, the head posture scoring module and the body movement scoring module, and to pre-process, analyze, count and record the received face recognition data, video surveillance data, head posture scoring data and body movement scoring data, and transmit the data analysis and processing results to the background management center; The face recognition module includes a face information collection camera for collecting face information, face recognition and automatic matching of student information; The head posture scoring module is used to obtain the student's head image by image recognition in the global coordinate system, calculate the head azimuth, and determine whether the yaw angle, roll angle, and pitch angle of the head at the current moment fall within the corresponding set range; The limb action scoring module uses human body morphology to perform human body action recognition and compares human body actions with standard actions; the human body morphology includes appearance, depth, optical flow and human skeleton; The obstacle unit is used to assess the enemy situation observation capability; The sprint unit is used to detect whether the examinee has reached the finish line and record the sprint time; The background management center is used to send the information of the examinees to the handheld terminal via the wireless network, receive the examinees' examination results transmitted by the handheld terminal, receive the video processing results transmitted by the video image processing unit, analyze and count the final examination results, and send the final examination results to the on-site real-time display terminal.

2. The intelligent system for individual tactical training and assessment according to claim 1 is characterized by: The video monitoring unit includes a low / high posture crawling monitoring camera, an obstacle monitoring camera and a sprint monitoring camera; The low / high posture crawling monitoring camera is used to obtain the original image video of the trainee's crawling posture during the period from low posture entry to high posture exit; The obstacle monitoring camera is used to obtain the original image video of the trainees' tactical actions using terrain and objects, and transmit the video to the data acquisition and analysis center to assess the trainees' enemy observation ability; The sprint monitoring camera is used to obtain original image videos of the trainees’ tactical actions in the sprint area; The video monitoring unit transmits the acquired original image video to the data acquisition and analysis center, and after being analyzed and processed by the algorithm of the data acquisition and analysis center, it is transmitted to the background management center; The terminal also includes at least one on-site real-time display terminal, which is used to display the assessment results; The movement posture monitoring module also includes at least two module cameras (201) for collecting images of the trainee's head.

3. The intelligent system for individual tactical training and assessment according to claim 2 is characterized by: Also includes NVR; The NVR is connected to the data acquisition and analysis center and the background management center respectively. The data acquisition and analysis center marks the original image and video clips, and the marked original image and video clips are transmitted to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image and video clips through the marked timestamp.

4. The intelligent system for individual tactical training and assessment according to claim 3 is characterized by: The head posture scoring module implements head posture scoring by the following steps: 1) Constructing a first backbone network model to collect and recognize the head image of the trainee in real time; the first backbone network model is a Darknet-53 backbone network; 1.1) The Darknet-53 backbone network includes a plurality of convolution blocks and fully connected layers connected sequentially from the input end to the output end; each convolution block includes a convolution layer connected sequentially, a batch normalization layer and an activation layer set in parallel; the convolution layer is composed of 1x1 and 3x3 convolution kernels; when the convolution step size is 1 by default, the padding is consistent by default; when the convolution step size is 2, the padding is valid; 1.2) Downsample the Darknet-53 backbone network with a sampling step of 1 or 2 each time; then concatenate the upsampled middle layer and the penultimate layer of the Darknet-53 backbone network to form a new optimized network, which is used to continuously iterate new sample training and learning to obtain the head model; 1.3) Optimizing and training the head model through a loss function; the loss function is one or more of target positioning offset loss, target confidence loss, and target classification loss; 1.4) Collect the head image information of the trainee to be tested, transmit it to the head model, and recognize the trainee's head image; 2) Calculate the head orientation angle 2.1) Constructing a second backbone network model, wherein the second backbone network model is a ResNet50 backbone network; the ResNet50 backbone network includes three fully connected layers, and the three fully connected layers correspond to the yaw angle, the roll angle, and the pitch angle of the head azimuth respectively; the mapping interval range of the ResNet50 backbone network is [-99, +99]; each fully connected layer makes a separate prediction within the mapping interval, and the fully connected numbers for separate predictions are all binary numbers; 2.2) Train the constructed ResNet50 backbone network to obtain the trained ResNet50 backbone network and obtain the actual labeled data; 2.3) The head image of the student identified in step 1.4) is transmitted to the ResNet50 backbone network trained in step 2.2) to calculate the head orientation angle; 2.3.1) Map the yaw angle, roll angle or pitch angle of the head azimuth in the student's head image to the interval [-99, +99] respectively, and classify the 199 values ​​in the mapping interval range from -99 to +99 in sequence, three in a group, and the middle interval as a discrete category; 2.3.2) The discrete categories obtained in step 2.3.1) are normalized into probability values ​​after being operated by the softmax function, and the probability values ​​of all discrete category data in the fully connected layer are added to 1; 2.3.3) Multiply the probability value obtained in step 2.3.2) by the corresponding discrete category and sum them to get the expected value; calculate the mean square error between the expected value and the angle of the actual labeled data, and finally output the head orientation angle through end-to-end training; 3) Determine whether the yaw angle, roll angle or pitch angle of the head azimuth at the current moment falls within the corresponding set range; if not, perform corresponding deduction operations according to the deduction rules and give a score; if so, give a score directly.

5. The intelligent system for individual tactical training and assessment according to claim 4 is characterized by: The body movement scoring module implements body movement scoring through the following steps: 1) Collect students’ human posture images in real time and identify target images; 1.1) Establish the first backbone network model; 1.2) Extract 8, 16, and 32 times downsampled features from the first backbone network model built in step 1.1); 1.3) Input the features extracted in step 1.2) with 8, 16, and 32 times downsampling into the feature pyramid network for multi-scale feature fusion; 1.4) Output the fusion result of step 1.3) to the detection head to detect and identify the target image; 2) Evaluate human posture; 2.1) Construct a lightweight network; the lightweight network consists of six layers of convolution: The first layer is the Conv / Relu layer, with a core size of 3×3, a stride of 2, and a channel of 32; The second layer is the Conv / Relu layer, with a core size of 3×3, a stride of 1, and 64 channels; The third layer is the Conv / Relu layer, with a core size of 3×3, a stride of 2, and 64 channels; The fourth layer is the Conv / Relu layer, with a core size of 3×3, a stride of 1, and 64 channels; The fifth layer is the TransConv / BN / Relu layer, with a core size of 2×2, a stride of 2, and 64 channels; The sixth layer is the TransConv / BN / Relu layer, with a core size of 2×2, a stride of 2, and a channel of 32; 2.2) Transferring the potential posture from a pre-trained teacher model to a constructed lightweight network to form a student model through a fast posture distillation model training method; the fast posture distillation model training method is to discriminate through a heat map of key bone nodes of the human body; 2.3) Downsample the target image detected and identified in step 1.4) twice through strided convolution; 2.4) The student model is trained using the latent posture learned in the teacher model, and the target image obtained by two downsamplings in step 2.3) is transmitted to the student model to obtain the confidence of the target image; the confidence of the target image is then compared with the preset confidence standard value. If it is greater than the confidence standard value, an image display of the heat map of the key bone nodes of the human body is output, and then the scoring system is entered and the score is output; if it is less than the confidence standard value, a corresponding point deduction operation is performed, and then the scoring system is entered and the score is output.

6. The intelligent system for individual tactical training and assessment according to claim 5 is characterized by: The frame structure of the low frame crawling module (1) is consistent with that of the high frame crawling module (3); The low frame crawling module (1) comprises a reinforcing rib assembly and legs, a connecting beam and a horizontal connecting seat arranged on the left and right sides; The supporting legs include a front leg (101), a middle leg (102) and a rear leg (103); the connecting beam includes a front beam (104) and a rear beam (105); the horizontal connecting seat includes a front seat (106), a middle seat (107) and a rear seat (108); wherein the top end of the front leg (101) is fixedly connected to the front end of the front seat (106), the top end of the middle leg (102) is fixedly connected to the middle of the middle seat (107), and the top end of the rear leg (103) is fixedly connected to the rear end of the rear seat (108); the rear end of the front seat (106) and the front end of the front beam (104), the rear end of the front beam (104) and the front end of the middle seat (107), the rear end of the middle seat (107) and the front end of the rear beam (105), and the rear end of the rear beam (105) and the front end of the rear seat (108) are respectively hinged via a connecting beam hinge shaft (115) and fixed via a positioning pin (116); The reinforcing rib assembly comprises three groups of connecting rods respectively arranged between two front seats (106) on the left and right sides, between two middle seats (107) and between two rear seats (108); each group of connecting rods comprises a hinged left rod (109) and a right rod (110), a left sleeve (111) and a middle sleeve (112) sleeved on both ends of the left rod (109), and a right sleeve (113) sleeved on the right end of the right rod (110); the left end of the left rod (109) is hinged to the horizontal connecting seat on the left side; the right end of the right rod (110) is hinged to the horizontal connecting seat on the right side; the left sleeve (111) is used to fix the left rod (109) is hinged to the left horizontal connecting seat; the middle sleeve (112) is used to fix the hinge of the left rod (109) and the right rod (110); the right sleeve (113) is used to fix the hinge of the right rod (110) and the right horizontal connecting seat; wherein the three hinge axes of the connecting rod between the two front seats (106) are parallel to the front seat (106); the three connecting beam hinge axes (115) of the connecting rod between the two middle seats (107) are parallel to the middle seat (107); the three connecting beam hinge axes (115) of the connecting rod between the two rear seats (108) are parallel to the rear seat (108); The number of the timing sensor groups (202) is three; the three timing sensor groups (202) are respectively arranged on the inner side surfaces of the two legs at the front end, the two legs at the change position, and the two legs at the rear end, and are used for timing the trainer's entry into the net, change of position, and exit from the net; the plurality of foul detection sensor groups (203) are respectively arranged on the inner side surfaces of the left and right connecting beams of each low-frame crawling module (1) and high-frame crawling module (3), and are used for foul action detection; the module cameras (201) are respectively arranged in the middle of a group of connecting rods at different positions, and are used for collecting images of the trainer's head; The acquisition control module comprises at least one acquisition control box (2) and a support air socket (204) arranged on the outer side of the connection beam of each low-frame crawling module (1) and high-frame crawling module (3); the acquisition control box (2) is electrically connected to the support air socket (204) via a cable; the support air socket (204) is electrically connected to a timing sensor group (202) and a foul detection sensor group (203) via a cable.

7. The intelligent system for individual tactical training and assessment according to claim 6 is characterized by: The sprint unit comprises two sprint rods (22), a collection control box (2), and at least one photoelectric sensor (23); Each sprint rod (22) is provided with a sprint rod navigation socket (24), or one of the sprint rods (22) is provided with two sprint rod navigation sockets (24); The photoelectric sensor (23) is arranged on the sprint pole (22) and is used to detect a signal indicating that the trainee has passed the finish line; The acquisition control box (2) is electrically connected to the sprint pole navigation socket (24) via a cable; The sprint rod navigation socket (24) is electrically connected to the photoelectric sensor (23) via a cable.

8. The intelligent system for individual tactical training and assessment according to claim 7 is characterized by: The comparison between the human body movement and the standard movement is achieved by calculating the Euclidean distance between the human body movement and the standard movement; The timing sensor group (202) and the foul detection sensor group (203) are both laser beam type photoelectric switch pairs; The two sprinting rods (22) are two hollow sprinting rods, each of which is higher than two meters; the upper ends of the two hollow sprinting rods are connected by a hollow fixed balancing rod (25); The photoelectric sensor (23) is a pair of laser beam-type photoelectric switches, and the receiving end and the transmitting end of the laser beam-type photoelectric switch are respectively mounted on two hollow sprint rods and are aligned toward each other; The crawling unit also includes a buzzer arranged in the acquisition control box (2); The handheld terminals include combat sports handheld terminals and terrain and object handheld terminals; The first and acquisition control boxes (2) are both provided with single-chip microcomputers, and all laser-beam type photoelectric switches are connected to the GPIO port of the single-chip microcomputer for signal transmission, and then the data is sent to the portable handheld terminal in a wireless manner.

9. A method for individual tactical training and assessment, based on the assessment intelligent system according to any one of claims 1 to 8, characterized in that: The following steps are involved: 1) Face recognition 1.1) Collect students’ facial information through facial information collection cameras and transmit them to the data collection and analysis center; 1.2) The data collection and analysis center will structure the student facial information collected in the early stage through face detection, face alignment, feature extraction, and feature matching, and upload the stored data to the backend management center; 1.3) Automatically identify and match student information, and once the match is successful, the student will enter the crawling unit assessment; 2) Crawling unit assessment 2.1) The acquisition control box (2) in the crawling unit enters the power-on initialization and self-test mode; 2.2) The acquisition control box (2) is connected to the combat sports handheld terminal via wireless. If the connection is abnormal, a buzzer alarm is triggered and the system returns to the power-on initialization and self-test mode. If the connection is normal, an infrared alignment test of the laser beam photoelectric switch pair is performed. If the test is abnormal, the status is sent to the combat sports handheld terminal and the system returns to the power-on initialization and self-test mode. If the test is normal, the system enters the test mode. 2.3) The backend management center sends the information of the students to be tested who have been successfully matched in step 1.3) to the combat sports handheld terminal through the wireless network, and the students enter the test waiting state; at the same time, the low / high posture crawling monitoring camera is turned on to obtain the original image video of the students crawling posture during the period from low posture network entry to high posture network exit; 2.4) Entering the crawling unit assessment mode, when the assessment trainee enters the network, the first timing sensor group (202) is triggered, and the acquisition control box (2) connected thereto starts timing, and at the same time the module camera (201) starts shooting; the acquisition control box (2) uploads the test signal to the combat sports handheld terminal; when the assessment trainee advances to the position where the posture is changed, the second timing sensor group (202) is triggered, and the acquisition control box (2) records the time of the posture change, and uploads the recorded time to the combat sports handheld terminal; when the assessment trainee exits the network, the third timing sensor group (202) is triggered, and the acquisition control box (2) records the time of exiting the network, and transmits the recorded time to the combat sports handheld terminal; once the assessment trainee touches the laser beam of the upper foul detection sensor group (203) during the crawling process, a foul behavior is generated, the corresponding indicator light on the acquisition control box (2) is lit, and a buzzer alarm is issued to remind the trainee to adjust the posture, and the foul information is transmitted to the combat sports handheld terminal; 2.5) After the assessment student leaves the network, the crawling unit assessment is completed, and the combat sports handheld terminal sends the assessment results to the background management center; the image captured by the module camera (201) is transmitted to the data collection and analysis center for analysis and processing, and then the data information processed by the data collection and analysis center is transmitted to the background management center; the original image video obtained by the low / high posture crawling monitoring camera is sent to the data collection and analysis center, and after analysis and processing, it is transmitted to the background management center; 3) Obstacle unit assessment 3.1) Enter the obstacle unit assessment and turn on the obstacle monitoring camera to obtain the original image video of the trainees' tactical actions using terrain and objects; 3.2) After the obstacle unit assessment is completed, the original image video obtained by the obstacle monitoring camera is transmitted to the data acquisition and analysis center, and then transmitted to the background management center after analysis and processing; 4) Sprint unit assessment 4.1) The acquisition control box (2) in the sprint unit enters the power-on initialization and self-test mode; 4.2) The acquisition control box (2) is connected to the terrain and object handheld terminal via wireless. If the connection is abnormal, a buzzer alarm is triggered and the system returns to the power-on initialization and self-test mode. If the connection is normal, the infrared alignment detection of the photoelectric sensor (23) is performed. If the detection is abnormal, the status is sent to the terrain and object handheld terminal and the system returns to the power-on initialization and self-test mode. If the detection is normal, the system enters the waiting state. 4.3) The backend management center sends the information of the students to be tested who have been successfully matched in step 1.3) to the terrain handheld terminal through the wireless network, and the students enter the test waiting state; at the same time, the sprint monitoring camera is turned on to obtain the original image video of the tactical actions of the students in the sprint area; 4.4) Enter the sprint unit assessment mode When the student reaches the end point, the infrared laser beam is blocked, and the acquisition control box (2) automatically records the current end time, and analyzes and judges the recorded current end time. If the result is unsatisfactory, a buzzer alarm sounds; at the same time, the recorded time information and alarm information are wirelessly transmitted to the terrain handheld terminal; 4.5) The terrain and object handheld terminal sends the assessment results to the background management center, and transmits the original image video obtained by the sprint monitoring camera to the data acquisition and analysis center, and then transmits it to the background management center after analysis and processing; 5) The backend management center analyzes and compiles the crawling unit assessment results, obstacle unit assessment results, and sprint unit assessment results of the examinees, and sends the final assessment results to the on-site real-time display terminal.

10. A method for individual tactical training and assessment according to claim 9, characterized in that: The step 2) also includes step 2.6); 2.6) The original image video acquired by the low / high posture crawling surveillance camera is sent to the data acquisition and analysis center, which marks the original image video clips and transmits the marked original image video clips to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, which can access and search the original image video clips through the marked timestamp; The step 3) also includes step 3.3); 3.3) The original image video acquired by the obstacle monitoring camera is transmitted to the data acquisition and analysis center, which marks the original image video clips and transmits the marked original image video clips to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, which can access and search the original image video clips through the marked timestamp; The step 4) also includes step 4.6); 4.6) The original image video obtained by the sprint monitoring camera is transmitted to the data acquisition and analysis center. The data acquisition and analysis center marks the original image video clips, and the marked original image video clips are transmitted to the NVR for recording and storage; at the same time, the data acquisition and analysis center transmits the marked timestamp to the background management center, and the background management center can access and search the original image video clips through the marked timestamp.

Images