A standing long jump landing performance detection method based on time series

Abstract

This invention discloses a time-series-based method for detecting the landing performance of a standing long jump. The method decomposes the long jump motion and generates scatter plots for each decomposed motion based on the time series. The landing situation is determined based on the scatter plots, and the method identifies the frame containing the landing moment in the long jump image where the jump distance can be accurately calculated. The final long jump distance is then calculated quickly and accurately. The calculation process does not require setting complex image processing thresholds, determining the jump foot posture, or identifying skeletal key points. It can handle more landing situations, has greater versatility, is simple to calculate, has a fast processing speed, and high accuracy. Furthermore, the long jump image is processed before calculation, segmenting the human body and background, minimizing interference from the background during the calculation of the segmented image.

Description

Technical Field

[0001] This invention relates to the field of standing long jump distance measurement technology, and in particular to a method for detecting the landing performance of standing long jump based on time series. Background Technology

[0002] In traditional standing long jump competitions, the landing point is typically determined manually before distance measurement. However, with rapid technological advancements and the trend towards intelligentization, machine vision is increasingly being used for distance measurement in standing long jump competitions, as disclosed in Chinese patent applications 202010052744.3 and 202111632521.5. A key aspect of machine vision-based distance measurement methods lies in determining the landing point, which directly determines and influences the accuracy and reliability of the measurement. Existing standing long jump distance measurement methods utilize the angle and threshold relationship between key points of the human skeleton formed between two consecutive image frames to determine whether the landing posture is foot or hip landing. Edge detection methods are then used to find the coordinates of the landing body part, thereby performing distance calculation. However, current methods for determining landing posture based on skeletal key points are quite complex. The uncertainty and complexity of athletes' landing postures make it impossible to use fixed skeletal angle thresholds for posture analysis, resulting in high computational difficulty. Furthermore, edge detection methods based on traditional image processing cannot effectively avoid interference caused by different styles and colors of shoes, shadows, ground noise, etc., leading to inaccurate landing point determination and large errors in the final standing jump score calculation. Summary of the Invention

[0003] The purpose of this invention is to provide a method for detecting the landing performance of the standing long jump based on time series.

[0004] The technical solution to achieve the objective of this invention is: a method for detecting the landing performance of the standing long jump based on time series, comprising the following steps:

[0005] S1. Use a camera to capture the entire long jump process from the side of the athlete, obtaining multiple long jump images;

[0006] S2. Process the long jump image. The processing includes: segmenting the human body and background in the long jump image, and treating the segmented human body as a unified whole.

[0007] S3. Establish an xy coordinate system, where the x-axis represents the horizontal distance traveled and the y-axis represents the height above the ground;

[0008] S4. Based on the long jump images, select the lowest pixel point P(x, y) of the human body in each long jump image;

[0009] S5. Constructing the time series S xand time series S y From the moment of preparation for takeoff until the athlete leaves the field, a time series S is generated. x Scatter plot and time series S y A scatter plot; where the time series S x The time series S is a sequence of values ​​along the x-axis of the lowest pixel P(x, y) of the human body in each frame of the long jump image, arranged in chronological order. y It is a sequence of values ​​of the lowest pixel point P(x,y) of the human body in each frame of the long jump image, arranged in chronological order along the y-axis.

[0010] S6. Based on the time series S x Scatter plot and time series S y The distribution of the scatter plot is used to determine the athlete's landing situation, and based on the athlete's landing situation, the frame of the long jump image where the landing time of the athlete can be accurately calculated after the long jump is completed is found.

[0011] S7. Based on the long jump image of the frame where the long jump distance can be accurately calculated after the athlete lands, find the value of the lowest pixel point P(x,y) of the body part in the x-axis direction, and calculate the final long jump distance score according to the ratio between the established xy coordinate system and the actual long jump field.

[0012] Furthermore, in step S6, the time series S x The value in the time series S increases with time. y The values ​​in the time series first increase and then decrease over time until they reach a minimum and remain constant. At this point, the athlete's landing is determined to be a non-backward landing, and the time series S is identified. y The frame containing the value of the long jump jump just as it drops to a constant value is used as the frame containing the landing moment after the athlete's long jump, allowing for accurate calculation of the jump distance. Based on the time series S... x Scatter plot and time series S y The distribution trend of the scatter plot can be used to determine the landing situation during the standing long jump; the landing situation during the standing long jump usually includes landing without backwards and landing backwards. Landing without backwards and landing backwards are reflected in the time series S. x and time series S y The difference in distribution trends is that when the landing does not regress, the time series S x The value in the time series S increases with time; when the time series is regressed, the value in the time series S increases with time. x The values ​​in the time series S first increase, then decrease, and then continue to increase. In other words, according to the time series S... x and time series S yThe distribution trend can be used to determine the landing situation during the standing long jump. Furthermore, different landing situations during the standing long jump correspond to different determinations of the landing point. If the horizontal movement distance of the lowest pixel point P(x, y) of the human body is the time series S... x The values ​​in the time series S increase with time, exhibiting a forward jump; the height of the lowest pixel P(x,y) above the ground in the human body part is the time series S. y The values ​​in the time series S all first increase and then decrease as time increases, until they reach a minimum value and remain constant, exhibiting a jump and fall pattern. At this point, according to the time series S... x and time series S y Based on the distribution trend, it can be determined that the landing situation during the standing long jump was without backward movement. Based on this landing situation, the time series S can be determined. y The long jump image of the frame where the value in time series S just dropped to remain constant (i.e., in time series S). y The image above (represented as the long jump image at the point where the parabola formed by connecting the values ​​is tangent to the time axis) is the long jump image of the frame where the landing time after the athlete's long jump can be accurately calculated. Thus, by finding the long jump image of the frame where the landing time after the athlete's long jump can be accurately calculated, the long jump distance can be accurately calculated.

[0013] Furthermore, in step S6, the time series S x The value in the time series S first increases, then decreases, and then continues to increase over time. y The values ​​in the time series first increase and then decrease over time until they reach a minimum and remain constant. At this point, the athlete's landing is considered to be backwards, and the time series S is identified. x The frame containing the moment when the value in the long jump drops from its highest point to its second rise is used as the frame containing the landing moment after the athlete's long jump, allowing for accurate calculation of the jump distance. For example, the horizontal movement distance of the lowest pixel point P(x, y) of the human body is represented by the time series S. x The value in the sequence increases, then decreases, and then continues to increase over time. This is manifested as jumping forward, then leaning backward, and then getting up again. At this point, it can be determined that the landing situation during the standing long jump is backward. Based on this landing situation, the frame that leans backward is determined, i.e., the time series S. x The long jump image of the frame where the value in time series S drops from its highest point and then rises again (i.e., in time series S). x The long jump image shown above (the frame containing the lowest point of the "V" shape) is the frame containing the landing moment when the athlete lands after the long jump, allowing for accurate calculation of the long jump distance. Thus, by finding the frame containing the landing moment when the athlete lands after the long jump, the long jump distance can be accurately calculated.

[0014] Furthermore, the camera mentioned in step S1 is a high-definition camera with a frame rate of 48 FPS or higher.

[0015] Furthermore, the camera mentioned in step S1 is positioned 3m to 4m to the side of the athlete during the long jump.

[0016] Furthermore, the camera mentioned in step S1 is positioned at a horizontal height of 1.2~1.8m.

[0017] Furthermore, the camera mentioned in step S1 is positioned at a horizontal height of 1.5m.

[0018] Furthermore, in step S2, the human body and background are segmented using the human body segmentation AI model Bodypix, the PASNet deep learning framework, the FCN deep learning framework, or the UET deep learning framework.

[0019] Furthermore, in step S3, the center point of the take-off line of the standing long jump is used as the origin of the established xy coordinate system. The origin of the xy coordinate system can be set arbitrarily, but compared with other methods, the calculation is simpler when the center point of the take-off line of the standing long jump is used as the origin of the established xy coordinate system.

[0020] The time-series-based method for detecting landing performance in the standing long jump has the following advantages:

[0021] 1. Before calculation, the long jump image is processed to segment the human body and the background. The segmented image is less affected by the background, and the detection can reduce the requirements of the background. No matter how complex the background is, it can be calculated accurately, which improves the accuracy of the landing point judgment under complex background and improves the accuracy of the standing long jump distance measurement.

[0022] 2. During detection, the segmented human body is treated as a unified whole, and the landing point is determined only by the trajectory of the lowest point P(x,y) of this unified whole, and is unrelated to the foot posture during the long jump. During detection, there is no need to determine the foot posture of the long jump or identify key skeletal points, which can handle more landing situations and has greater versatility.

[0023] 3. This invention decomposes the long jump motion and forms scatter plots of each decomposed motion based on the time series. The landing situation is determined based on the scatter plot of the time series, and the long jump image of the frame where the landing time can be accurately calculated is found. The final long jump distance score is calculated quickly and accurately. The calculation process does not require setting complex image processing thresholds. The calculation is simple, fast, and highly accurate.

[0024] 4. In the standing long jump landing performance detection method based on time series of the present invention, after presenting the movement trajectory in the form of time series scatter plot, not only can the entire long jump action be seen more clearly and intuitively, but it also provides relevant data such as height, distance and time for subsequent data analysis of the standing long jump, which is convenient for correcting the long jump action and improving the long jump performance.

[0025] 5. Based on the pure image processing method of a monocular camera, the equipment is simple and low in cost. It has no specific requirements for the placement of the camera and does not require additional camera calibration. The equipment is easy to install at the long jump test site. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the process structure of the standing long jump landing performance detection method based on time series according to the present invention;

[0027] Figure 2 This is a schematic diagram of the camera installation structure in the time-series-based standing long jump landing performance detection method of the present invention;

[0028] Figure 3 The time series S is the result obtained by measuring the landing performance of the standing long jump using the time series-based method of this invention when the landing situation is such that the jump does not involve backward movement. x Scatter plot;

[0029] Figure 4 The time series S is obtained by measuring the landing performance of the standing long jump using the time series-based method of this invention. y Scatter plot;

[0030] Figure 5 The time series S is obtained by measuring the landing performance of the standing long jump using the time series-based method of this invention when the landing situation is backward during the long jump. x Scatter plot. Implementation

[0031] The preferred embodiment of the time-series-based standing long jump landing performance detection method of the present invention will be described in detail below with reference to the accompanying drawings:

[0032] like Figure 1 As shown, a time-series-based method for detecting landing performance in the standing long jump includes the following steps:

[0033] S1. Install camera 1 at a position approximately 3m-4m to the side of the athlete during the long jump, at a horizontal height of about 1.5m. Figure 2 As shown, the camera is a high-definition camera with a frame rate of 48 FPS or higher; the camera is used to capture the entire long jump process from the side of the athlete, resulting in multiple long jump images;

[0034] S2. Process the long jump image. The processing includes: using the human body segmentation AI model Bodypix to segment the human body and background in the long jump image, and treating the segmented human body as a unified whole.

[0035] S3. Establish an xy coordinate system, where the x-axis represents the horizontal movement distance, the y-axis represents the height above the ground, and the units for both the x and y axes are pixels. The center point of the take-off line for the standing long jump is used as the origin of the established xy coordinate system.

[0036] S4. Based on the long jump images, select the lowest pixel point P(x, y) of the human body in each long jump image;

[0037] S5. Constructing the time series S x and time series S y From the moment of preparation for takeoff until the athlete leaves the field, a time series S is generated. x Scatter plot and time series S y A scatter plot; where the time series S x The time series S is a sequence of values ​​along the x-axis of the lowest pixel P(x, y) of the human body in each frame of the long jump image, arranged in chronological order. y It is a sequence of values ​​of the lowest pixel point P(x,y) of the human body in each frame of the long jump image, arranged in chronological order along the y-axis.

[0038] S6. Based on the time series S x Scatter plot and time series S y The distribution of the scatter plot is used to determine the athlete's landing situation, and based on the athlete's landing situation, the frame containing the landing moment after the athlete's long jump landing is located, allowing for accurate calculation of the long jump distance; specifically including:

[0039] If the time series S x The value in increases with time, such as Figure 3 As shown, time series S y The value in the table first increases and then decreases over time, until it reaches a minimum value and remains constant, such as... Figure 4 As shown, at this point, the athlete's landing is determined to be a non-backward landing, and the time series S is found. y The frame containing the value of the long jump is the one where the value just drops to a constant value. This frame is then used as the frame containing the landing moment after the athlete lands, at which the jump distance can be accurately calculated.

[0040] If the time series S x The value in [the graph] first increases, then decreases, and then continues to increase over time, such as [example]. Figure 5 As shown, time series S y The value in the table first increases and then decreases over time, until it reaches a minimum value and remains constant, such as... Figure 4 As shown, at this point, the athlete's landing situation is determined to be a backward landing, and the time series S is found. x The frame containing the value that drops from its highest point to rise again is the long jump image. This long jump image is then used as the frame containing the landing moment after the athlete lands, allowing for accurate calculation of the long jump distance.

[0041] S7. Based on the long jump image of the frame where the long jump distance can be accurately calculated after the athlete lands, find the value of the lowest pixel point P(x,y) of the body part in the x-axis direction, and calculate the final long jump distance score according to the ratio between the established xy coordinate system and the actual long jump field.

[0042] This invention relates to a time-series-based method for detecting the landing performance of a standing long jump. In practice, camera 1 is positioned to the side of the athlete during the long jump, filming the entire process from the athlete's side. Once the athlete is ready, the long jump distance measurement program is initiated, and a start command is issued. The camera begins recording the athlete's standing long jump. Within a certain timeframe after the command is issued, the athlete completes the standing long jump, and camera 1 stops recording, transmitting each captured frame to a backend server. The backend server processes each frame and calculates the final long jump distance.

[0043] The time-series-based method for detecting landing performance in the standing long jump has the following advantages:

[0044] 1. Before calculation, the long jump image is processed to segment the human body and the background. The segmented image is less affected by the background, and the detection can reduce the requirements of the background. No matter how complex the background is, it can be calculated accurately, which improves the accuracy of the landing point judgment under complex background and improves the accuracy of the standing long jump distance measurement.

[0045] 2. During detection, the segmented human body is treated as a unified whole, and the landing point is determined only by the trajectory of the lowest pixel P(x,y) of this unified whole, and is unrelated to the foot posture during the long jump. During detection, there is no need to determine the foot posture of the long jump or identify key skeletal points, which can handle more landing situations and has stronger versatility.

[0046] 3. This invention decomposes the long jump motion and forms scatter plots of each decomposed motion based on the time series. The landing situation is determined based on the scatter plot of the time series, and the long jump image of the frame where the landing time can be accurately calculated is found. The final long jump distance score is calculated quickly and accurately. The calculation process does not require setting complex image processing thresholds. The calculation is simple, fast, and highly accurate.

[0047] 4. In the standing long jump landing performance detection method based on time series of the present invention, after presenting the movement trajectory in the form of time series scatter plot, not only can the entire long jump action be seen more clearly and intuitively, but it also provides relevant data such as height, distance and time for subsequent data analysis of the standing long jump, which is convenient for correcting the long jump action and improving the long jump performance.

[0048] 5. Based on the pure image processing method of a monocular camera, the equipment is simple and low in cost. It has no specific requirements for the placement of the camera and does not require additional camera calibration. The equipment is easy to install at the long jump test site.

[0049] This invention provides a time-series-based method for detecting standing long jump landing performance. In step S6, based on the time series S... x Scatter plot and time series S y The distribution trend of the scatter plot can be used to determine the landing situation during the standing long jump; the landing situation during the standing long jump usually includes landing without backwards and landing backwards. Landing without backwards and landing backwards are reflected in the time series S. x and time series S y The difference in distribution trends is that when the landing does not regress, the time series S x The value in the time series S increases with time; when the time series is regressed, the value in the time series S increases with time. x The values ​​in the time series S first increase, then decrease, and then continue to increase. In other words, according to the time series S... x and time series S y The distribution trend can be used to determine the landing situation during the standing long jump. At the same time, different landing situations during the standing long jump also correspond to different rulings of the landing point.

[0050] The present invention provides a method for detecting the landing performance of the standing long jump based on time series. In step S6, specifically, if the time series S... x The value in the time series S increases with time. y The values ​​in the time series first increase and then decrease over time until they reach a minimum and remain constant. At this point, the athlete's landing is determined to be a non-backward landing, and the time series S is identified. yThe frame containing the moment when the value in the long jump just drops to a constant value is used as the frame containing the landing moment after the athlete's long jump, allowing for accurate calculation of the jump distance. The horizontal movement distance of the lowest pixel point P(x,y) of the human body is the time series S. x The values ​​in the time series S increase with time, exhibiting a forward jump; the height of the lowest pixel P(x,y) above the ground in the human body part is the time series S. y The values ​​in the time series S all first increase and then decrease as time increases, until they reach a minimum value and remain constant, exhibiting a jump and fall pattern. At this point, according to the time series S... x and time series S y Based on the distribution trend, it can be determined that the landing situation during the standing long jump was without backward movement. Based on this landing situation, the time series S can be determined. y The long jump image of the frame where the value in time series S just dropped to remain constant (i.e., in time series S). y The image above (represented as the long jump image at the point where the parabola formed by connecting the values ​​is tangent to the time axis) is the long jump image of the frame where the landing time after the athlete's long jump can be accurately calculated. Thus, by finding the long jump image of the frame where the landing time after the athlete's long jump can be accurately calculated, the long jump distance can be accurately calculated.

[0051] The present invention provides a method for detecting the landing performance of the standing long jump based on time series. In step S6, specifically, if the time series S... x The value in the time series S first increases, then decreases, and then continues to increase over time. y The values ​​in the time series first increase and then decrease over time until they reach a minimum and remain constant. At this point, the athlete's landing is considered to be backwards, and the time series S is identified. x The frame containing the moment when the value in the long jump drops from its highest point to its second rise is used as the frame containing the landing moment after the athlete's long jump, allowing for accurate calculation of the jump distance. For example, the horizontal movement distance of the lowest pixel point P(x, y) of the human body is represented by the time series S. x The value in the sequence increases, then decreases, and then continues to increase over time. This is manifested as jumping forward, then leaning backward, and then getting up again. At this point, it can be determined that the landing situation during the standing long jump is backward. Based on this landing situation, the frame that leans backward is determined, i.e., the time series S. x The long jump image of the frame where the value in time series S drops from its highest point and then rises again (i.e., in time series S). xThe long jump image shown above (the frame containing the lowest point of the "V" shape) is the frame containing the landing moment when the athlete lands after the long jump, allowing for accurate calculation of the long jump distance. Thus, by finding the frame containing the landing moment when the athlete lands after the long jump, the long jump distance can be accurately calculated.

[0052] The present invention provides a time-series-based method for detecting the landing performance of the standing long jump. Preferably, the camera mentioned in step S1 is a high-definition camera with a frame rate of 48 FPS or higher.

[0053] The present invention provides a time-series-based method for detecting the landing performance of the standing long jump. Preferably, in step S1, the camera is positioned 3m to 4m to the side of the athlete during the long jump.

[0054] The present invention provides a time-series-based method for detecting the landing performance of the standing long jump. Preferably, in step S1, the camera is positioned at a horizontal height of 1.2 to 1.8 meters.

[0055] The present invention provides a time-series-based method for detecting the landing performance of the standing long jump. Preferably, in step S1, the camera is positioned at a horizontal height of 1.5m.

[0056] The present invention provides a time-series-based method for detecting the landing performance of the standing long jump. In step S2, in addition to using the Bodypix AI model for human body segmentation to segment the human body and background, the PASNet deep learning framework, FCN deep learning framework, or UET deep learning framework can also be used to segment the human body and background.

[0057] The present invention provides a time-series-based method for detecting the landing performance of the standing long jump. Preferably, in step S3, the center point of the take-off line of the standing long jump is used as the origin of the established xy coordinate system. The origin of the xy coordinate system can be arbitrarily set, but compared with other methods, the calculation is simpler when the center point of the take-off line of the standing long jump is used as the origin of the established xy coordinate system.

[0058] The method for detecting the landing performance of standing long jump based on time series is presented in this invention. For those skilled in the art, there are several simple deductions or substitutions that can be made without departing from the concept of this invention, and all such deductions or substitutions should be considered to fall within the scope of protection of this invention.

Claims

1. A time series-based standing long jump landing performance detection method, characterized in that: Includes the following steps: S1. Use a camera to capture the entire long jump process from the side of the athlete, obtaining multiple long jump images; S2. Process the long jump image. The processing includes: segmenting the human body and background in the long jump image, and treating the segmented human body as a unified whole. S3. Establish an xy coordinate system, where the x-axis represents the horizontal movement distance and the y-axis represents the height above the ground; S4. Based on the long jump images, select the lowest pixel point P(x, y) of the human body in each long jump image; S5. Construct time series Sx and time series Sy, starting from the start of the jump and ending when the athlete leaves the field, and generate scatter plots of time series Sx and time series Sy; where time series Sx is a sequence of values ​​of the lowest pixel point P(x,y) of the human body in each frame of the long jump image arranged in chronological order along the x-axis, and time series Sy is a sequence of values ​​of the lowest pixel point P(x,y) of the human body in each frame of the long jump image arranged in chronological order along the y-axis. S6. Based on the distribution of the scatter plots of time series Sx and time series Sy, determine the athlete's landing situation, and based on the athlete's landing situation, find the frame of the long jump image where the landing time of the athlete can be accurately calculated after the long jump landing. S7. Based on the long jump image of the frame where the long jump distance can be accurately calculated after the athlete lands, find the value of the lowest pixel point P(x,y) of the body part in the x-axis direction, and calculate the final long jump distance score according to the ratio between the established xy coordinate system and the actual long jump field. 2.The time series based standing long jump landing performance detection method according to claim 1, characterized in that: In step S6, the value in time series Sx increases with time, while the value in time series Sy first increases and then decreases with time until it reaches a minimum value that remains constant. At this point, it is determined that the athlete's landing situation is that the athlete did not fall backwards. The long jump image of the frame where the value in time series Sy just decreased to a constant value is found is used as the long jump image of the frame where the athlete lands after the long jump, and the landing time can be accurately calculated. 3.The time series based standing long jump landing performance detection method according to claim 1, characterized in that: In step S6, the value in time series Sx first increases and then decreases as time increases, and then continues to increase. The value in time series Sy first increases and then decreases as time increases until it reaches a minimum value that remains constant. At this point, the athlete's landing situation is determined to be a landing backward, and the frame containing the moment when the value in time series Sx drops from its highest point and then rises again is found. This found long jump image is used as the frame containing the landing moment after the athlete lands, at which the long jump distance can be accurately calculated. 4.The time series based standing long jump landing performance detection method according to claim 1, characterized in that: The camera mentioned in step S1 is a high-definition camera with a frame rate of 48 FPS or higher. 5.The time series based standing long jump landing performance detection method according to claim 1, characterized in that: The camera mentioned in step S1 is located 3m to 4m to the side of the athlete during the long jump.

6. The method for detecting the landing performance of the standing long jump based on time series as described in claim 1, characterized in that: The camera mentioned in step S1 is positioned at a horizontal height of 1.2~1.8m.

7. The method for detecting the landing performance of the standing long jump based on time series as described in claim 6, characterized in that: The camera mentioned in step S1 is positioned at a horizontal height of 1.5m.

8. The method for detecting the landing performance of the standing long jump based on time series as described in claim 1, characterized in that: In step S2, the Bodypix AI model for human body segmentation is used to segment the human body and the background.

9. The method for detecting the landing performance of the standing long jump based on time series as described in claim 1, characterized in that: In step S2, the human body and background are segmented using the PASNet deep learning framework, the FCN deep learning framework, or the UET deep learning framework.

10. The method for detecting the landing performance of the standing long jump based on time series as described in claim 1, characterized in that: In step S3, the center point of the take-off line of the standing long jump is used as the origin of the established xy coordinate system.

Images