AI-based sports examination room system and sports examination scoring method

Abstract

The invention relates to the technical field of control management systems, in particular to an AI-based sports examination room system and a sports examination scoring method.The AI-based sports examination room system comprises a plurality of cameras and further comprises an area dividing unit, a track is divided into a plurality of unit recognition areas, and a starting area and a terminal area are divided by combining the distribution positions of a starting line and a terminal line of the track; the preparation identification unit is used for acquiring image data in different starting areas, analyzing and calculating to obtain a target starting line, and marking the starting area corresponding to the target starting line as a preparation area; the project confirmation unit is used for determining a plurality of to-be-determined terminal lines and to-be-determined end point areas corresponding to the preparation areas, analyzing the average speed of the testers passing through the to-be-determined end point areas and the average speed of the testers passing through the two unit identification areas before and after the to-be-determined end point areas, comparing and analyzing the speeds of the testers before and after passing through the to-be-determined terminal lines, and determining a target terminal line; and determining a test item based on the target finishing line and the target starting line.

Description

Technical Field

[0001] The present invention relates to the technical field of control management systems, and in particular to an AI-based sports examination system and a sports examination scoring method. Background Art

[0002] The AI-based sports examination system is designed to automatically identify, manage and evaluate various aspects of sports examinations through computer vision, deep learning and sensor technology. The system accurately identifies the identity, movement type and posture quality of the examinee, monitors the completion of running, long jump, throwing and other projects in real time and generates personalized performance reports. By integrating artificial intelligence into the sports examination, schools can improve the objectivity of assessment standards, provide students with more scientific guidance, and thus promote the overall improvement of physical education.

[0003] When testing candidates in running events, the physical examination system in the prior art often requires staff to actively input the event content and adjust the corresponding monitoring equipment for monitoring in order to obtain the running data of the candidates during the running test. It is difficult to automatically identify the test items and intelligently control the equipment to monitor and record the test data, which affects the overall efficiency and automation of the system. Summary of the invention

[0004] In view of the above-mentioned shortcomings of the prior art, the present invention provides an AI-based sports examination system and a sports examination scoring method, which can effectively solve the problem of low degree of automation in running event identification, monitoring and management in the prior art.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0006] The present invention provides an AI-based sports examination room system, which at least includes: a plurality of cameras, and also includes:

[0007] The area division unit divides the runway into multiple unit identification areas, divides the start area and the finish area based on the distribution of the start line and the finish line of the runway, and constructs area pairs based on the start line and finish line corresponding to different events;

[0008] The preparation recognition unit obtains image data in different starting areas, recognizes the human body contour in the monitoring image based on the human body contour recognition algorithm, analyzes and calculates the target starting line according to the distribution of the human body contour and the position of the starting line, and records the starting area corresponding to the target starting line as the preparation area;

[0009] The project confirmation unit determines multiple pending finish lines and pending finish areas corresponding to the preparation area through the area pair group, obtains the instantaneous speed of each human body contour in the preparation area, and when the instantaneous speed of each human body contour changes from 0 to a positive number at the same time, marks each human body contour as a test person and screens the pending finish line;

[0010] Analyze the average speed of the test personnel passing through the pending finish area and the two unit identification areas before and after the pending finish area, and compare and analyze the speeds of the test personnel before and after passing each pending finish line to determine the target finish line, and determine the test items based on the target finish line and the target starting line.

[0011] Furthermore, the division and binding process of the area division unit for the starting area and the finish area is as follows:

[0012] A single runway picture is obtained and recorded as a unit picture, the unit pictures are spliced ​​to form a complete picture, the area of ​​the runway region in each unit picture is obtained and recorded as the unit runway area, the unit runway area is divided by the runway width to obtain the unit coverage length, and the minimum value of the unit coverage length is recorded as the minimum division length;

[0013] The runway is divided into multiple unit identification areas, and the division of the unit identification areas satisfies:

[0014] The starting line or finish line of any event is within the same unit identification area, while the starting line or finish line of different events may not be within the same unit identification area;

[0015] The inner side length of the unit recognition area is less than or equal to the minimum division length;

[0016] The inner length of the unit recognition area is greater than or equal to the length threshold. The length threshold calculation formula is L′=t0*v0, where:

[0017] t0 is the preset shortest recognition time;

[0018] v0 is the maximum running speed of humans;

[0019] The unit identification area containing the starting line is recorded as the starting area, and the unit identification area containing the finish line is recorded as the finish area. Area pairs are constructed according to the starting lines and finish lines corresponding to different events, and each area pair includes a starting area and a finish area.

[0020] Furthermore, the steps of the preparation identification unit identifying the preparation area and the target starting line are as follows:

[0021] PART1: The benchmark images in each starting area are preset, and the real-time images in each starting area are obtained as monitoring images. The monitoring images are compared with the benchmark images, and the ratio of the area of ​​different areas between the monitoring images and the benchmark images to the total area of ​​the images is calculated to obtain the difference ratio. A ratio threshold is preset, and when the difference ratio is greater than , enter PART2;

[0022] PART2: The human body contour in the monitoring image is identified and marked by the human body contour recognition algorithm. The center point of the human body contour is used as the position point of each human body contour. Each position point is marked in the monitoring image and recorded as P i , where i = 1, 2, 3, ..., j, j is the total number of human body contours, calculate the distance between any two position points and select the farthest distance as the first judgment parameter;

[0023] PART3: Mark each starting line in the monitoring image, calculate the proximity reference value between each position point and different starting lines through the proximity reference value calculation formula, select the minimum proximity reference value as the second judgment parameter, and the proximity reference value calculation formula is:

[0024] in:

[0025] d i is the shortest distance between each position point and the starting line;

[0026] j is the total number of human body contours;

[0027] Δd is the adjacent reference value corresponding to the starting line;

[0028] PART4: The first reference value and the second reference value are normalized and then input into the starting reference value calculation formula to calculate the starting reference value. When the starting reference value is less than or equal to the preset starting threshold, the starting line corresponding to the second reference value is used as the target starting line, and the corresponding starting area is recorded as the preparation area. The image in the preparation area is continuously analyzed. The starting reference value calculation formula is:

[0029] α=λ1*|L0-β1|+λ2*β2, where:

[0030] λ1 and λ2 are both preset weight coefficients;

[0031] L0 is the width of the runway;

[0032] β1 and β2 are the first reference value and the second reference value respectively;

[0033] α is the starting reference value.

[0034] Furthermore, the process of the item confirmation unit identifying the test item is as follows:

[0035] S1: Periodically collect high-definition images in the preparation area. The collection period of high-definition images is greater than or equal to the preset reaction time. A unique label is assigned to each human body contour in the preparation area. The unique label is bound to the face data corresponding to each human body contour. The current instantaneous speed of each human body contour is obtained. When the instantaneous speed of each human body contour changes from 0 to a positive number at the same time, the corresponding moment is recorded as the starting moment, and each human body contour is marked as a test person.

[0036] S2: obtaining multiple finish lines corresponding to the target starting line as pending finish lines, recording the finish area corresponding to the pending finish line as the pending finish area, recording the two unit identification areas before and after each pending finish area as the pre-finish area area and the post-finish area area of ​​the pending finish area, and performing S3 for each pending finish area and its pre-finish area and post-finish area;

[0037] S3: Get the average passing speed in the three areas and record them as The average passing speed is the average speed of all testers passing through each area. When a tester does not pass through an area, the average passing speed in that area is recorded as 0. satisfy and When , proceed to S4;

[0038] S4: Obtain each pending finish line in the pending finish area and analyze them one by one. The independent analysis process is as follows:

[0039] The average speed of the tester within T time before passing the undetermined finish line is recorded as the first judgment speed v1′, and the average speed of the tester within T time after passing the undetermined finish line is recorded as the second judgment speed v′2. v = v′2-v1′ to calculate the sprint judgment value β v , when the sprint judgment value is less than or equal to the preset sprint judgment threshold, the tester is recorded as a sprinter;

[0040] When the proportion of sprinters among testers is greater than the preset personnel proportion threshold, the pending finish line is recorded as the target finish line, and the area to which the target finish line belongs is recorded as the target finish area;

[0041] S5: Obtain a target finish line and a target starting line, and determine the test items of the current tester based on the target finish line and the target starting line.

[0042] Furthermore, a dynamic monitoring unit is also included, specifically:

[0043] The area passing array is determined based on the route corresponding to the test item. The area passing array is composed of multiple unit identification areas. The number of times each tester passes in the unit identification area is recorded and combined with the passing records of the testers in the unit identification area, the first section area and the last section area in the area passing array are marked, and video monitoring is performed on the first section area, the last section area in the area passing array and the unit identification area between the two.

[0044] Furthermore, the video monitoring process of the dynamic monitoring unit is as follows:

[0045] The unit recognition area is marked as Q n , where n is the serial number of the unit identification area, n = 1, 2, 3, ..., m, m is the number of unit identification areas, obtain the passing route corresponding to the test item, determine the area passing array based on the passing route, the area passing array is an ordered array, and the elements in the area passing array are the passing routes that pass through the unit identification areas in sequence;

[0046] Recording unit identification area Q n The number of times each tester passes the test, combined with the unit identification area Q n The testers in the Q n Mark as The marking process is as follows:

[0047] q is the maximum number of times each tester passes;

[0048] p is the minimum number of times each tester passes;

[0049] Get the value corresponding to the maximum value of q Recorded as Where a is the corresponding unit identification area number, x is the maximum value of q, and p is the minimum value corresponding to Terminal Area Where b is the corresponding unit identification area number, y is the maximum value of p, and the selected area passes through the Q that appears for the xth time in the array. a Recorded as the first section, select the area passing through the yth occurrence of Q in the array b The area passing through the first section area, the last section area and the unit identification area between the two in the array is recorded as the monitoring area, and the video image in the monitoring area is collected.

[0050] Furthermore, the number of testers in different monitoring areas is obtained, and a unit number threshold is preset. When the number of testers in the same monitoring area is greater than the unit number threshold, the monitoring area is recorded as a key monitoring area, and the video image resolution and video frame rate collected in the key monitoring area are increased.

[0051] A sports test scoring method based on AI, comprising the following steps:

[0052] Step 1: Divide the runway into multiple unit identification areas, record the unit identification areas containing the starting line or the finish line as the starting area and the finish area respectively, bind the starting line and the finish line belonging to the same running event, and construct an area pair group containing the starting area and the finish area;

[0053] Step 2: Obtain image data in different starting areas, identify the human body contour in the monitoring image based on the human body contour recognition algorithm, analyze the distribution of the human body contour and the position of the starting line, and when multiple human body contours are evenly distributed along the starting line, record the starting line as the target starting line, and record the starting area corresponding to the target starting line as the preparation area;

[0054] Step 3: Determine multiple pending finish lines and pending finish areas corresponding to the preparation area through area pair grouping, obtain the instantaneous speed of each human body contour in the preparation area, and when the instantaneous speed of each human body contour changes from 0 to a positive number at the same time, mark each human body contour as a test person and screen the pending finish lines;

[0055] Analyze the average speed of the tester passing through the pending finish area and the two unit identification areas before and after the pending finish area, and compare and analyze the speed of the tester before and after passing each pending finish line to determine the target finish line, and determine the test items based on the target finish line and the target starting line;

[0056] Step 4: Determine the area passing array based on the route corresponding to the test item. The area passing array is composed of multiple unit identification areas. Record the number of times each tester passes in the unit identification area and combine it with the passing records of the testers in the unit identification area to mark the first and last areas in the area passing array. Perform video monitoring on the first and last areas in the area passing array and the unit identification areas between the two.

[0057] Compared with the known prior art, the technical solution provided by the present invention has the following beneficial effects:

[0058] 1. The area division unit of the present invention can divide the corresponding multiple groups of starting areas and finish areas based on different running events, so as to determine the test items to be performed by the test personnel according to the image data in the starting area and the finish area. The preparation identification unit can preliminarily judge whether the person on the track is in the preparation stage of the running test through the starting reference value, and then mark the running area as the preparation area. By identifying the preparation area, the starting area can be continuously monitored before the start of the running test, so as to facilitate the identification of the starting action of the test personnel in the subsequent process, and make further analysis based on the starting action, monitor the overall process of the running event, and obtain complete running data of each test personnel.

[0059] 2. The project confirmation unit of the present invention first determines multiple pending finish lines by screening, thereby narrowing the confirmation scope, so as to more quickly find the target finish line corresponding to the target starting line, and the speed change characteristics of the test person before and after passing the target finish line can be identified through speed analysis. Generally, the test person will accelerate and sprint before passing the target finish line, and reduce stopping after passing the target finish line. Based on the above speed characteristics, the target finish line among the pending finish lines can be quickly confirmed, which is conducive to confirming the running project currently being performed by the test person, and can realize intelligent self-service identification without the need for staff input and adjustment, and has a high level of automation.

[0060] 3. The dynamic monitoring unit of the present invention can determine the monitoring area based on the test items and the location data of each tester. By collecting video images in the monitoring area, it can observe the running process of all testers while reducing the image capture of other irrelevant cameras, thereby reducing the bandwidth occupancy of video uploading, and can reduce the number of cameras in use, thereby reducing power loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0062] Figure 1 It is the overall module block diagram of the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] The present invention will be further described below in conjunction with the embodiments.

[0065] When conducting different running tests, different subject test contents are usually included. In this embodiment, the 50m, 100m, 800m and 1000m running test items commonly seen in life are taken as the main identification and analysis objects to specifically describe the specific implementation methods of the present invention.

[0066] See also Figure 1 , an AI-based sports examination system, comprising at least: a plurality of cameras arranged outside a runway, for capturing image information on the runway, the shooting range of the plurality of cameras covering the entire runway, so that the running test persons can be recorded by the cameras at any position on the runway.

[0067] The area division unit divides the runway into multiple unit areas for identification based on different screen coverage of multiple cameras, and divides the starting area and the finishing area into the distribution positions of the runway starting line and the finishing line (the starting line and the finishing line in the present invention refer to the starting line and the finishing line of each event, and the starting line and the finishing line are not necessarily a continuous line segment, but may also include multiple line segments) and binds them, wherein:

[0068] The single runway images captured by each camera are recorded as unit images, and the complete runway images are recorded as complete images. The runway area in each unit image is obtained and its area is calculated to obtain the unit runway area. The unit runway area is divided by the runway width to obtain the unit coverage length. The minimum value of the unit coverage length corresponding to each unit image is recorded as the minimum division length. The runway is divided into multiple unit recognition areas. The division of the unit recognition area satisfies:

[0069] The starting line or finish line of any event is within the same unit identification area (so that testers on the same starting line or passing the same finish line can be within the same unit identification area), and the starting lines or finish lines of different events may not be within the same unit identification area;

[0070] The inner side length of the unit identification area is less than or equal to the minimum division length (the inner side refers to the inner side length of the unit identification area), so that the runway image within the unit identification area can be completely captured;

[0071] The inner length of the unit recognition area is greater than or equal to the length threshold. The length threshold calculation formula is L′=t0*v0, where:

[0072] t0 is the preset shortest recognition time (i.e., the moving speed of the test person can be recognized through the video of this time. In a specific implementation, the shortest recognition time is 5 seconds);

[0073] v0 is the maximum running speed of humans (in a specific embodiment, the value is 10 meters per second);

[0074] The unit identification area containing the starting line is recorded as the starting area, and the unit identification area containing the finish line is recorded as the finish area. Area pairs are constructed according to the starting lines and finish lines corresponding to different items. Each area pair contains a starting area and a finish area (that is, through the area pair, one or more finish areas can be determined under the premise of knowing a starting area, and further, a test item can be determined through each set of starting lines and finish lines)

[0075] The preparation recognition unit analyzes the runway image data recorded by cameras in different starting areas to obtain the preparation area. The preparation area represents the start preparation stage of the running test for the personnel in the starting area, where:

[0076] PART1: The benchmark images in each starting area are preset. The benchmark image refers to the image collected without any obstruction in the starting area. The real-time image in each starting area is obtained as the monitoring image. The monitoring image is compared with the benchmark image. The ratio of the area of ​​different areas between the monitoring image and the benchmark image to the total area of ​​the image is calculated to obtain the difference ratio. A ratio threshold is preset. When the difference ratio is greater than , enter PART2;

[0077] PART2: The human body contour in the monitoring image is identified and marked by the human body contour recognition algorithm. The center point of the human body contour is used as the position point of each human body contour. Each position point is marked in the monitoring image and recorded as P i , where i = 1, 2, 3, ..., j, j is the total number of human body contours, calculate the distance between any two position points and select the farthest distance as the first judgment parameter (the distance between the position points can be obtained by multiplying the pixel distance between the two position points by the characteristic proportional coefficient, and the first judgment parameter reflects the distance between the two farthest persons on the runway);

[0078] It should be noted that the existing technology includes a variety of different AI human contour recognition algorithms. Common human contour recognition algorithms include convolutional neural networks (YOLO, Faster R-CNN), edge detection algorithms (Canny edge detection, Sobel and Laplacian edge detection) and instance segmentation algorithms (Mask R-CNN, DeepLab series), etc. The above algorithms can accurately identify the human contour in the image and display the contour.

[0079] PART3: Mark each starting line in the monitoring image (the same starting line refers to the starting line of the same event), and calculate the proximity reference value between each position point and different starting lines through the proximity reference value calculation formula (the proximity reference value can reflect the average distance between the people on the runway and each starting line. When the average distance approaches 0, it can be determined that the people on the runway are evenly arranged along the starting line). Select the minimum proximity reference value as the second judgment parameter. The proximity reference value calculation formula is:

[0080] in:

[0081] d i is the shortest distance between each position point and the starting line;

[0082] j is the total number of human body contours;

[0083] Δd is the adjacent reference value corresponding to the starting line;

[0084] PART4: The first reference value and the second reference value are normalized and then input into the starting reference value calculation formula to calculate the starting reference value. When the starting reference value is less than or equal to the preset starting threshold, the starting line corresponding to the second reference value is used as the target starting line, and the corresponding starting area is recorded as the preparation area. The image in the preparation area is continuously analyzed. The starting reference value calculation formula is:

[0085] α=λ1*|L0-β1|+λ2*β2, where:

[0086] λ1 and λ2 are both preset weight coefficients;

[0087] L0 is the width of the runway;

[0088] β1 and β2 are the first reference value and the second reference value respectively;

[0089] α is the starting reference value;

[0090] The more evenly the people on the track are distributed along the starting line, and the closer the maximum distance between any two people is to the width of the track, the smaller the starting reference value will be. When the test personnel enter the preparation stage of the running test, the above two data characteristics will be met, which will also lead to a smaller starting reference value obtained by analysis. This means that the starting reference value can be used to preliminarily determine whether the people on the track are in the preparation stage of the running test, and then mark the running area as the preparation area. By identifying the preparation area, the starting area can be continuously monitored before the start of the running test, which makes it easier to identify the starting action of the test personnel in the subsequent process, and make further analysis based on the starting action, monitor the overall process of the running project, and obtain complete running data for each test personnel.

[0091] The project confirmation unit identifies the starting action of the person in the preparation area and marks the person in the area as a test person. Based on the speed at which the test person passes through the finish line in different finish areas, the finish area corresponding to the running test project currently being conducted is selected, thereby identifying the current test project, wherein:

[0092] S1: Periodically collect high-definition images in the preparation area, and the collection period of the high-definition images is greater than or equal to the preset reaction time (in a specific embodiment, the reaction time is the maximum value of the human reaction time range, i.e., 0.3 seconds). Identify each human body contour in the high-definition image and assign a unique label to each human body contour. The unique label is bound to the facial data corresponding to each human body contour, so that each human body contour can be identified and distinguished by the unique label. Since the unique label is bound to the facial data, different human body contours will not be confused. Compare the currently collected high-definition image with the most recently collected high-definition image to obtain the displacement distance of the human body contour, and then obtain the current instantaneous speed of each human body contour (the instantaneous speed is equal to the displacement distance divided by the collection period). When the instantaneous speed of each human body contour changes from 0 to a positive number at the same time, the corresponding moment is recorded as the starting moment, and each human body contour is marked as a tester;

[0093] S2: Obtain multiple finish lines corresponding to the target starting line (i.e., the starting line and finish line of the same running event) as pending finish lines, record the finish area corresponding to the pending finish line as pending finish area, record the two unit identification areas before and after each pending finish area as the pre-finish area area and the post-finish area area of ​​the pending finish area (the concept of front and back is based on the order of the test personnel passing through, and the area before the pending finish area is the pre-finish area area), and perform S3 for each pending finish area and its pre-finish area area and post-finish area area;

[0094] S3: Get the average passing speed in the three areas and record them as They represent the average passing speed in the area before the end point, the area to be determined, and the area after the end point. The average passing speed is the average speed of all test personnel passing through each area. The average speed is equal to the average of all instantaneous speeds. When a test person does not pass through an area, the average passing speed in the area is recorded as 0. satisfy and When , proceed to S4;

[0095] S4: Obtain each pending finish line in the pending finish area and analyze them one by one. The independent analysis process is as follows:

[0096] The average speed of the tester within T time before passing the undetermined finish line is recorded as the first judgment speed v1′, and the average speed of the tester within T time after passing the undetermined finish line is recorded as the second judgment speed v′2. v = v′2-v1′ to calculate the sprint judgment value β v , when the sprint judgment value is less than or equal to the preset sprint judgment threshold, the tester is recorded as a sprinter;

[0097] When the proportion of sprinters among testers is greater than the preset personnel proportion threshold, the pending finish line is recorded as the target finish line, and the area to which the target finish line belongs is recorded as the target finish area;

[0098] S5: Obtain a target finish line and a target starting line, and determine the test item of the current tester based on the target finish line and the target starting line, that is, the running item of the current test.

[0099] It should be noted that multiple pending finish lines are first determined by screening, thereby narrowing the confirmation scope, so as to more quickly find the target finish line corresponding to the target starting line, and through the speed analysis of steps S3 and S4, the speed change characteristics of the test person before and after passing the target finish line can be identified. Generally, the test person will accelerate and sprint before passing the target finish line, and reduce stopping after passing the target finish line. Based on the above speed characteristics, the target finish line among the pending finish lines can be quickly confirmed, which is conducive to confirming the running project currently being performed by the test person, and can realize intelligent self-service identification without the need for staff input and adjustment, and has a high level of automation.

[0100] The dynamic monitoring unit adjusts the collection within each unit's identification area based on the test items and the location data of each tester, including:

[0101] The unit recognition area is marked as Q n , where n is the serial number of the unit identification area, n = 1, 2, 3, ..., m, m is the number of unit identification areas, obtain the passing route corresponding to the test item, determine the area passing array based on the passing route, the area passing array is an ordered array, and the elements in the area passing array are the passing routes passing through the unit identification areas in sequence (the area passing array may contain repeated elements);

[0102] Recording unit identification area Q n The number of times each tester passes the test, combined with the unit identification area Q n The testers in the Q n Mark as The marking process is as follows:

[0103] q is the maximum number of times each tester passes (i.e., the number of times each tester passes Q n times, and take the maximum value);

[0104] p is the minimum number of times each tester passes;

[0105] Get the value corresponding to the maximum value of q Recorded as Where a is the corresponding unit identification area number, x is the maximum value of q, and p is the minimum value corresponding to Terminal Area Where b is the corresponding unit identification area number, y is the maximum value of p, and the selected area passes through the Q that appears for the xth time in the array. a Recorded as the first section, select the area passing through the yth occurrence of Q in the array b The area passing through the first area, the last area and the unit identification area between the two in the array is recorded as the monitoring area, and multiple cameras corresponding to the monitoring area are started to collect video images in the monitoring area to record the movement of the test personnel.

[0106] It should be noted that by collecting video images in the monitoring area, it is possible to observe the running process of all test personnel while reducing the image capture of other irrelevant cameras, thereby reducing the bandwidth occupancy of video uploads, and can reduce the number of cameras in use, thereby reducing power loss.

[0107] Furthermore, the number of testers in different monitoring areas is obtained, and a unit number threshold is preset. When the number of testers in the same monitoring area is greater than the unit number threshold, the monitoring area is recorded as a key monitoring area, and the video image resolution and video frame rate collected in the key monitoring area are increased, so that when the testers violate the rules or a collision occurs, the video image can be slowed down or enlarged for refereeing.

[0108] A sports test scoring method based on AI, comprising the following steps:

[0109] Step 1: Divide the runway into multiple unit identification areas, record the unit identification areas containing the starting line or the finish line as the starting area and the finish area respectively, bind the starting line and the finish line belonging to the same running event, and construct an area pair group containing the starting area and the finish area;

[0110] Step 2: Obtain image data in different starting areas, identify the human body contour in the monitoring image based on the human body contour recognition algorithm, analyze the distribution of the human body contour and the position of the starting line, and when multiple human body contours are evenly distributed along the starting line, record the starting line as the target starting line, and record the starting area corresponding to the target starting line as the preparation area;

[0111] When the human body contours are evenly distributed along the starting line, it can be explained that the distribution characteristics of the human body contours meet the distribution characteristics of the starting preparation stage of the running event, thereby determining the preparation area and the target starting line.

[0112] Step 3: Determine multiple pending finish lines and pending finish areas corresponding to the preparation area through area pair grouping, obtain the instantaneous speed of each human body contour in the preparation area, and when the instantaneous speed of each human body contour changes from 0 to a positive number at the same time, mark each human body contour as a test person and screen the pending finish lines;

[0113] It should be noted that when the instantaneous speed of the human body contour changes from 0 to a positive number at the same time, it is considered to meet the characteristic requirements of the starting phase of the running event, so that the test person is determined and the test item is determined based on the test person's movement.

[0114] Analyze the average speed of the tester passing through the pending finish area and the two unit identification areas before and after the pending finish area, and compare and analyze the speed of the tester before and after passing each pending finish line to determine the target finish line, and determine the test items based on the target finish line and the target starting line;

[0115] Step 4: Determine the area passing array based on the route corresponding to the test item. The area passing array is composed of multiple unit identification areas. Record the number of times each tester passes in the unit identification area and combine it with the passing records of the testers in the unit identification area to mark the first and last areas in the area passing array. Perform video monitoring on the first and last areas in the area passing array and the unit identification areas between the two.

[0116] Through the above steps, AI intelligent recognition technology can be used to quickly determine the test items currently being conducted on the track and observe and record the testers, which helps to analyze the relevant data of different testers when conducting running tests and facilitates the scoring of the testers' running ability through image data.

[0117] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sports examination room system based on AI, comprising a plurality of cameras, characterized in that: Also includes: The area division unit divides the runway into multiple unit identification areas, divides the start area and the finish area based on the distribution of the start line and the finish line of the runway, and constructs area pairs based on the start line and finish line corresponding to different events; The preparation recognition unit obtains image data in different starting areas, recognizes the human body contour in the monitoring image based on the human body contour recognition algorithm, analyzes and calculates the target starting line according to the distribution of the human body contour and the position of the starting line, and records the starting area corresponding to the target starting line as the preparation area; The project confirmation unit determines multiple pending finish lines and pending finish areas corresponding to the preparation area through the area pair group, obtains the instantaneous speed of each human body contour in the preparation area, and when the instantaneous speed of each human body contour changes from 0 to a positive number at the same time, marks each human body contour as a test person and screens the pending finish line; Analyze the average speed of the test personnel passing through the pending finish area and the two unit identification areas before and after the pending finish area, and compare and analyze the speeds of the test personnel before and after passing each pending finish line to determine the target finish line, and determine the test items based on the target finish line and the target starting line.

2. The AI-based sports examination system according to claim 1, characterized in that: The process of dividing and binding the starting area and the finishing area by the area division unit is as follows: A single runway picture is obtained and recorded as a unit picture, the unit pictures are spliced ​​to form a complete picture, the area of ​​the runway region in each unit picture is obtained and recorded as the unit runway area, the unit runway area is divided by the runway width to obtain the unit coverage length, and the minimum value of the unit coverage length is recorded as the minimum division length; The runway is divided into multiple unit identification areas, and the division of the unit identification areas satisfies: The starting line or finish line of any event is within the same unit identification area, while the starting line or finish line of different events may not be within the same unit identification area; The inner side length of the unit recognition area is less than or equal to the minimum division length; The inner length of the unit recognition area is greater than or equal to the length threshold. The length threshold calculation formula is L′=t0*v0, where: t0 is the preset shortest recognition time; v0 is the maximum running speed of humans; The unit identification area containing the starting line is recorded as the starting area, and the unit identification area containing the finish line is recorded as the finish area. Area pairs are constructed according to the starting lines and finish lines corresponding to different events, and each area pair includes a starting area and a finish area.

3. The AI-based sports examination system according to claim 1, characterized in that: The steps of the preparation identification unit identifying the preparation area and the target starting line are as follows: PART1: The benchmark images in each starting area are preset, and the real-time images in each starting area are obtained as monitoring images. The monitoring images are compared with the benchmark images, and the ratio of the area of ​​different areas between the monitoring images and the benchmark images to the total area of ​​the images is calculated to obtain the difference ratio. A ratio threshold is preset, and when the difference ratio is greater than , enter PART2; PART2: The human body contour in the monitoring image is identified and marked by the human body contour recognition algorithm. The center point of the human body contour is used as the position point of each human body contour. Each position point is marked in the monitoring image and recorded as P i , where i = 1, 2, 3, ..., j, j is the total number of human body contours, calculate the distance between any two position points and select the farthest distance as the first judgment parameter; PART3: Mark each starting line in the monitoring image, calculate the proximity reference value between each position point and different starting lines through the proximity reference value calculation formula, select the minimum proximity reference value as the second judgment parameter, and the proximity reference value calculation formula is: in: d i is the shortest distance between each position point and the starting line; j is the total number of human body contours; Δd is the adjacent reference value corresponding to the starting line; PART4: The first reference value and the second reference value are normalized and then input into the starting reference value calculation formula to calculate the starting reference value. When the starting reference value is less than or equal to the preset starting threshold, the starting line corresponding to the second reference value is used as the target starting line, and the corresponding starting area is recorded as the preparation area. The image in the preparation area is continuously analyzed. The starting reference value calculation formula is: α=λ1*|L0-β1|+λ2*β2, where: λ1 and λ2 are both preset weight coefficients; L0 is the width of the runway; β1 and β2 are the first reference value and the second reference value respectively; α is the starting reference value.

4. The AI-based physical education examination system according to claim 3, characterized in that: The process of identifying the test items by the item confirmation unit is as follows: S1: Periodically collect high-definition images in the preparation area. The collection period of high-definition images is greater than or equal to the preset reaction time. A unique label is assigned to each human body contour in the preparation area. The unique label is bound to the face data corresponding to each human body contour. The current instantaneous speed of each human body contour is obtained. When the instantaneous speed of each human body contour changes from 0 to a positive number at the same time, the corresponding moment is recorded as the starting moment, and each human body contour is marked as a test person. S2: obtaining multiple finish lines corresponding to the target starting line as pending finish lines, recording the finish area corresponding to the pending finish line as the pending finish area, recording the two unit identification areas before and after each pending finish area as the pre-finish area area and the post-finish area area of ​​the pending finish area, and performing S3 for each pending finish area and its pre-finish area and post-finish area; S3: Get the average passing speed in the three areas and record them as The average passing speed is the average speed of all testers passing through each area. When a tester does not pass through an area, the average passing speed in that area is recorded as 0. satisfy and When , proceed to S4; S4: Obtain each pending finish line in the pending finish area and analyze them one by one. The independent analysis process is as follows: The average speed of the tester within T time before passing the undetermined finish line is recorded as the first judgment speed v1′, and the average speed of the tester within T time after passing the undetermined finish line is recorded as the second judgment speed v′2. v = v′2-v1′ to calculate the sprint judgment value β v , when the sprint judgment value is less than or equal to the preset sprint judgment threshold, the tester is recorded as a sprinter; When the proportion of sprinters among testers is greater than the preset personnel proportion threshold, the pending finish line is recorded as the target finish line, and the area to which the target finish line belongs is recorded as the target finish area; S5: Obtain a target finish line and a target starting line, and determine the test items of the current tester based on the target finish line and the target starting line.

5. The AI-based sports examination system according to claim 4, characterized in that: It also includes a dynamic monitoring unit for video monitoring, specifically: The area passing array is determined based on the route corresponding to the test item. The area passing array is composed of multiple unit identification areas. The number of times each tester passes in the unit identification area is recorded and combined with the passing records of the testers in the unit identification area, the first section area and the last section area in the area passing array are marked, and video monitoring is performed on the first section area, the last section area in the area passing array and the unit identification area between the two.

6. The AI-based physical education examination system according to claim 5, characterized in that: The video monitoring process of the dynamic monitoring unit is as follows: The unit recognition area is marked as Q n , where n is the serial number of the unit identification area, n = 1, 2, 3, ..., m, m is the number of unit identification areas, obtain the passing route corresponding to the test item, determine the area passing array based on the passing route, the area passing array is an ordered array, and the elements in the area passing array are the passing routes that pass through the unit identification areas in sequence; Recording unit identification area Q n The number of times each tester passes the test, combined with the unit identification area Q n The testers in the Q n Mark as The marking process is as follows: q is the maximum number of times each tester passes; p is the minimum number of times each tester passes; Get the value corresponding to the maximum value of q Recorded as Where a is the corresponding unit identification area number, x is the maximum value of q, and p is the minimum value corresponding to Terminal Area Where b is the corresponding unit identification area number, y is the maximum value of p, and the selected area passes through the Q that appears for the xth time in the array. a Recorded as the first section, select the area passing through the yth occurrence of Q in the array b The area passing through the first section area, the last section area and the unit identification area between the two in the array is recorded as the monitoring area, and the video image in the monitoring area is collected.

7. The AI-based physical education examination system according to claim 6, characterized in that: The number of testers in different monitoring areas is obtained, and a unit number threshold is preset. When the number of testers in the same monitoring area is greater than the unit number threshold, the monitoring area is recorded as a key monitoring area, and the video image resolution and video frame rate collected in the key monitoring area are increased.

8. An AI-based physical education examination scoring method, applied to an AI-based physical education examination system as described in claims 1-7, characterized in that: The following steps are involved: Step 1: Divide the runway into multiple unit identification areas, record the unit identification areas containing the starting line or the finish line as the starting area and the finish area respectively, bind the starting line and the finish line belonging to the same running event, and construct an area pair group containing the starting area and the finish area; Step 2: Obtain image data in different starting areas, identify the human body contour in the monitoring image based on the human body contour recognition algorithm, analyze the distribution of the human body contour and the position of the starting line, and when multiple human body contours are evenly distributed along the starting line, record the starting line as the target starting line, and record the starting area corresponding to the target starting line as the preparation area; Step 3: Determine multiple pending finish lines and pending finish areas corresponding to the preparation area through area pair grouping, obtain the instantaneous speed of each human body contour in the preparation area, and when the instantaneous speed of each human body contour changes from 0 to a positive number at the same time, mark each human body contour as a test person and screen the pending finish lines; Analyze the average speed of the tester passing through the pending finish area and the two unit identification areas before and after the pending finish area, and compare and analyze the speed of the tester before and after passing each pending finish line to determine the target finish line, and determine the test items based on the target finish line and the target starting line; Step 4: Determine the area passing array based on the route corresponding to the test item. The area passing array is composed of multiple unit identification areas. Record the number of times each tester passes in the unit identification area and combine it with the passing records of the testers in the unit identification area to mark the first and last areas in the area passing array. Perform video monitoring on the first and last areas in the area passing array and the unit identification areas between the two.

Images