Badminton falling point prediction method and system based on computer vision

Abstract

The invention relates to the technical field of visual inspection, in particular to a shuttlecock falling point prediction method and system based on computer vision, and the method comprises the following steps: capturing an image sequence at the hitting moment of a shuttlecock through a camera disposed on a shuttlecock court, extracting the position coordinates of the shuttlecock in the image sequence, and generating position coordinate data; and according to the position coordinate data, calculating the speed and direction of the shuttlecock, and generating the dynamic characteristics of the shuttlecock. According to the method, the image sequence of the hitting moment of the badminton is captured through the camera arranged on the badminton court, the position coordinate data of the badminton are extracted from the image sequence, and basic data are provided for predicting the falling point of the badminton. Based on the position coordinate data, the speed and direction of the badminton are calculated, and the dynamic characteristics of the badminton are further generated. The extraction of the dynamic information improves the dynamic responsiveness and the analysis depth of data processing.

Description

Technical Field

[0001] The present invention relates to the field of visual detection technology, and in particular to a method and system for predicting the landing point of a badminton based on computer vision. Background Art

[0002] Visual inspection technology, or machine vision technology, mainly involves the application of computer vision and image processing technology to automatically detect, analyze and process image data. The badminton landing prediction method is an application of visual inspection technology in sports technology, which is specially optimized for badminton. This method uses a camera installed above the badminton court to capture the flight trajectory of the ball and use image sequence analysis to predict the landing point of the badminton.

[0003] However, existing visual detection technology in badminton motion analysis usually relies on image processing and computer vision algorithms to analyze the position and trajectory of moving objects, but ignores real-time changes in environmental factors such as wind speed and wind direction. Due to the lack of comprehensive analysis of environmental factors, it is difficult to accurately reflect the actual flight trajectory of badminton, resulting in deviations in prediction results in practical applications. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a badminton landing point prediction method and system based on computer vision.

[0005] In order to achieve the above object, the present invention adopts the following technical scheme, a badminton landing point prediction method based on computer vision, comprising the following steps: Capturing an image sequence of a badminton hitting moment through a camera deployed on a badminton court, extracting the badminton position coordinates in the image sequence, and generating position coordinate data; calculating the speed and direction of the badminton based on the position coordinate data, and generating a dynamic feature of the badminton; Based on the dynamic characteristics of the badminton, a time series analysis is used to predict the flight trajectory of the badminton, and the expected position of the badminton at several time points in the future is calculated to generate a preliminary trajectory prediction result; according to the preliminary trajectory prediction result and the wind speed data obtained in real time from the wind speed and direction sensors around the venue, the predicted trajectory is adjusted to generate an adjusted trajectory prediction result; Using the adjusted trajectory prediction result, the landing area of ​​the badminton is calculated through a badminton landing point prediction algorithm to generate predicted landing point area data; the predicted landing point area data is compared with a venue area map to verify the landing point and generate a precise landing point verification result.

[0006] Preferably, the steps of acquiring the position coordinate data are: The camera deployed on the badminton court is used to capture the image sequence of the moment when the badminton is hit, and the original image data is obtained; Based on the original image data, the displacement of the badminton between two frames in the image sequence is extracted, and the calculation formula is:

[0007] in, is the coordinate of the badminton in the previous frame, is the coordinate of the badminton in the current frame, are the velocity components of the badminton in the x and y directions, is the displacement of the badminton between two frames; Based on the displacement of the badminton between the two frames, position coordinate data is obtained.

[0008] Preferably, the steps of acquiring the dynamic characteristics of the badminton are: Based on the position coordinate data, the speed and direction of the badminton are calculated using the following formula:

[0009] and

[0010] in, is the displacement in the x direction, is the displacement in the y direction, is the time interval, For speed, is the moving direction angle; Based on the calculated speed and direction, the dynamic characteristics of the badminton are summarized.

[0011] Preferably, the steps for obtaining the preliminary trajectory prediction result are: Based on the dynamic characteristics of the badminton, the movement trend of the badminton is analyzed using time series, the change pattern is evaluated, and a preliminary flight trajectory prediction result is obtained; Based on the preliminary flight trajectory prediction results, using multiple linear regression analysis, the expected position of the badminton at a future time point is calculated to obtain future position prediction data; Based on the future position prediction data, the prediction data at each time point are summarized and compared, and all prediction points are integrated to form a preliminary trajectory prediction result.

[0012] Preferably, the steps for obtaining the adjusted trajectory prediction result are: Based on the preliminary trajectory prediction results, collect wind speed and direction sensor data around the badminton court to obtain real-time wind speed and direction data; Based on the real-time wind speed and wind direction data, the badminton flight trajectory is recalculated to obtain a new trajectory point. The calculation formula is: ; in, is the original trajectory point, is the wind speed, is the speed of badminton, is the wind direction angle, is the angle of the badminton's flight direction, is the time interval, is the new trajectory point; Based on the new trajectory point, an adjusted trajectory prediction result is generated.

[0013] Preferably, the steps of acquiring the predicted landing area data are: Based on the adjusted trajectory prediction result, analyzing the potential landing point of the badminton in the venue to obtain the predicted landing point area coordinates; Based on the predicted coordinates of the landing area, the center and boundary of the landing area are calculated to obtain the center coordinates of the landing area. The calculation formula is: ; in, To predict the number of points in the landing area, is the weight assigned according to the density of the landing points, For the coordinates of each point, is the center coordinate of the weighted landing area; Based on the center coordinates of the landing point area, the range of the badminton landing point area is determined to form predicted landing point area data.

[0014] Preferably, the steps for obtaining the precise verification result of the landing point are: Collect the predicted landing point area data to obtain the coordinates and boundary information of the badminton landing point; Compare with the site area map to verify whether each landing point is within the site area and obtain the comparison result; Based on the comparison results, the position of each landing point is verified to determine whether it complies with the game rules and expectations, and an accurate landing point verification result is generated.

[0015] The present invention provides a badminton landing point prediction system, comprising: An image capture module is used to capture an image sequence of the moment of hitting the badminton, extract the position coordinates of the badminton in each frame, and generate position coordinate data; The dynamic analysis module is used to calculate the displacement and time difference between every two frames of the badminton according to the position coordinate data, deduce the speed and movement direction of the badminton, and obtain the dynamic characteristics of the badminton; The trajectory prediction module is used to predict the flight trajectory of the badminton based on the dynamic characteristics of the badminton by using time series analysis to obtain the position of the badminton at a future time point, and then adjust the flight trajectory in combination with the real-time wind speed data provided by the wind speed and direction sensors around the venue to obtain the adjusted trajectory prediction result; A landing point calculation module is used to calculate the landing point area of ​​the badminton using the adjusted trajectory prediction result and generate predicted landing point area data; The landing point verification module is used to compare the predicted landing point area data with the venue area map to verify whether the landing point of the badminton is accurate and generate an accurate landing point verification result.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: The present invention captures the image sequence of the hitting moment of the badminton through the camera deployed on the badminton court, and extracts the badminton position coordinate data therefrom, providing basic data for the prediction of the badminton landing point. Based on these position coordinate data, the speed and direction of the badminton are calculated, and the badminton dynamic characteristics are further generated. The extraction of this dynamic information improves the dynamic responsiveness and analysis depth of data processing. Then, the flight trajectory of the badminton is predicted according to the dynamic characteristics by using time series analysis, and the predicted trajectory is adjusted in combination with the wind speed data obtained from the wind speed and direction sensors around the venue in real time. This method of combining real-time environmental monitoring data not only improves the accuracy of the prediction, but also increases the adaptability to environmental changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] See also Figure 1 The present invention provides a technical solution, a method for predicting the landing point of a badminton ball based on computer vision, comprising the following steps: The camera deployed on the badminton court captures the image sequence of the moment when the badminton is hit, extracts the position coordinates of the badminton in the image sequence, and generates the position coordinate data; according to the position coordinate data, the speed and direction of the badminton are calculated to generate the dynamic characteristics of the badminton; Based on the dynamic characteristics of badminton, time series analysis is used to predict the flight trajectory of the badminton, and the expected position of the badminton at several time points in the future is calculated to generate a preliminary trajectory prediction result; according to the preliminary trajectory prediction result and the wind speed data obtained in real time from the wind speed and direction sensors around the venue, the predicted trajectory is adjusted to generate an adjusted trajectory prediction result; The adjusted trajectory prediction results are used to calculate the landing area of ​​the badminton and generate predicted landing area data; the predicted landing area data is compared with the venue area map to verify the landing point and generate accurate landing point verification results.

[0020] The steps to obtain the position coordinate data are: The camera deployed on the badminton court is used to capture the image sequence of the moment when the badminton is hit, and the original image data is obtained; Based on the original image data, the displacement of the badminton between two frames in the image sequence is extracted, and the calculation formula is: ; in, is the coordinate of the badminton in the previous frame, is the coordinate of the badminton in the current frame, are the velocity components of the badminton in the x and y directions, is the displacement of the badminton between two frames; Based on the displacement of the badminton between two frames, the position coordinate data is obtained.

[0021] Specifically, by recording the camera coordinates and corresponding lens calibration parameter information in the specific number and position index file of multiple cameras distributed on the badminton court, the execution process of the image sequence capturing the moment of badminton hitting is refined. First, according to the exposure time and aperture size data provided in the camera model manual, the parameters are matched with the pre-registered light sensor value and frame rate configuration table. The camera is controlled to access the image acquisition card and compared with the pixel clarity parameters recorded internally. When the pixel clarity exceeds the threshold value obtained by on-site measurement (the threshold value is derived from the average value obtained by statistics of the clarity of multiple images taken under different lighting conditions under actual venue conditions), the frame is captured and a series of consecutive frames of image sequences are cached in the memory.

[0022] The benefit of the formula is that the difference between the coordinates of the two frames is projected using the velocity direction, thereby obtaining the displacement along the velocity direction within the corresponding time interval, ensuring that effective displacement values ​​are extracted from complex motion trajectories.

[0023] The parameter acquisition step is to record the position of the badminton in the previous frame through the camera and use the calibrated pixel-to-meter conversion ratio to extract the pixel coordinates of the center point of the badminton image frame by frame, and then correspond to the position in the field calibration coordinate system to obtain ; The parameter acquisition step is to use the same camera and the same pixel coordinates to convert the calibration parameters to the actual coordinates in the next adjacent frame, and extract the corresponding position of the badminton to obtain , this value is determined by repeated measurements and taking the average; The parameter acquisition step is to obtain the velocity component by first-order difference based on multiple frames of continuous coordinate data and divide by time interval, and reduce the error by high frame rate recording, so as to obtain The speed range is derived from the statistical data of the flying speed of badmintons hit by professional players in actual competitions. The median value within a reasonable range is taken as the parameter after multiple field tests. Calculation process: Substitute the parameters into the formula: ; Calculate the numerator part:

[0024] Denominator: ; Overall calculation: ; The result shows that the displacement between the two frames along the current motion direction is about 0.0883 meters.

[0025] The steps to obtain the dynamic features of badminton are: Based on the position coordinate data, the speed and direction of the badminton are calculated using the following formula:

[0026] and

[0027] in, is the displacement in the x direction, is the displacement in the y direction, is the time interval, For speed, is the moving direction angle; Based on the calculated speed and direction, the dynamic characteristics of the badminton are summarized.

[0028] Specifically, the formula is beneficial in that it quantifies the displacement between two frames and obtains key motion features such as speed and direction, so that subsequent analysis steps can perform more in-depth motion trajectory prediction and feature extraction based on accurate speed and direction information.

[0029] The parameter acquisition step is to obtain the difference in the x-coordinate of the badminton between two adjacent frames by positioning the multiple frames of images in the previous link. In the previous step, it has been determined that the difference in the coordinates of the badminton between two consecutive frames is ; The parameters are obtained in the same way, from the difference between the y coordinates of the badminton in two adjacent frames of images. The above detection step records ; The parameter acquisition step is to determine it using the camera frame rate. When the camera frame rate is 240 frames per second, the time interval between each frame is ; Substitute the parameters into the formula to calculate the speed and direction of the badminton: First calculate the speed:

[0030] Bring in :

[0031] Then calculate the direction angle:

[0032] The result shows that the speed of the badminton between the current two frames is about 21.2 meters per second, and the direction angle is about 19.3 degrees. When the speed exceeds 20 meters per second, it means that the badminton is flying faster.

[0033] Through the speed and direction data recorded in the previous link, the speed interval classification table and direction angle grouping standard registered in the reference file are used to query and identify the speed value and direction value, and the speed value is matched with the speed interval divided step by step according to the measured ball speed range (10-30m / s) in the corresponding classification table. When the speed value falls into a specific interval, the corresponding interval identifier is stored in the internal registration unit, and the direction value is divided into angle segments through the established angle mapping table. The mapping table is based on the direction angle range generated when actually measuring various hitting methods. For example, the direction value below 10 degrees is marked as a horizontal segment, and 10 to 30 degrees is marked as an upward segment. When the direction value is within the specified range, the corresponding direction segment value is marked in the internal data record item, and the processed speed and direction mark values ​​are merged with the existing historical indicator table into a set of dynamic feature data arrays. The final generated dynamic feature data is stored in the specified data output file in a standard format through internal recording instructions.

[0034] The steps to obtain the preliminary trajectory prediction results are: Based on the dynamic characteristics of badminton, the movement trend of badminton is analyzed by time series, the change pattern is evaluated, and the preliminary flight trajectory prediction results are obtained; Based on the preliminary flight trajectory prediction results, multiple linear regression analysis is used to calculate the expected position of the badminton at a future time point to obtain the future position prediction data; Based on the future position prediction data, the prediction data at each time point are summarized and compared, and all prediction points are integrated to form a preliminary trajectory prediction result.

[0035] Specifically, based on the badminton dynamic feature data set, first identify the speed and direction changes of each data point, record the speed vector and movement direction of each time point in detail, and evaluate the acceleration of the badminton by calculating the displacement difference and time difference between two consecutive data points. Based on this, conduct an in-depth analysis of the movement trend and determine the main influencing factors of badminton flight, such as the aerodynamic drag coefficient and the influence of gravity acceleration. Apply these parameters to adjust the time series prediction model to achieve accurate prediction of the badminton's flight path in the short term. Through this iterative process, gradually optimize the model parameters until the prediction error reaches an acceptable safety range. Usually, the error threshold is set to no more than 5% to ensure the accuracy of the prediction, forming a preliminary flight trajectory prediction result. Based on the preliminary flight trajectory prediction results, multiple linear regression analysis is used to independently process the badminton position at each predetermined time point, combining the time variable and the predicted position variable to determine the expected position of the badminton at a specific time point in the future. The prediction value at each time point is checked for accuracy to ensure that the calculation of each prediction point is based on the prediction results of the previous time point. This method is used to comprehensively obtain future position prediction data to provide a prediction for the badminton flight trajectory; Using the future position prediction data, the badminton position at each time point is comprehensively analyzed, these data points are integrated, the position data at each time point are compared, and the consistency and reliability of the data are detected. This process not only involves the aggregation of data points, but also includes pattern recognition and error analysis of flight trajectories to ensure the accuracy of the entire predicted trajectory. In particular, the nonlinear characteristics of badminton motion are taken into account when processing each prediction point, and dynamic time bending is adopted to optimize the trajectory prediction results to generate preliminary trajectory prediction results.

[0036] The steps to obtain the adjusted trajectory prediction results are: Based on the preliminary trajectory prediction results, the wind speed and direction sensor data around the badminton court are collected to obtain real-time wind speed and direction data; Based on the real-time wind speed and direction data, the badminton flight trajectory is recalculated to obtain a new trajectory point. The calculation formula is:

[0037] in, is the original trajectory point, is the wind speed, is the speed of badminton, is the wind direction angle, is the angle of the badminton's flight direction, is the time interval, is the new trajectory point; Based on the new trajectory points, an adjusted trajectory prediction result is generated.

[0038] Specifically, the preliminary trajectory prediction results have been obtained from the previous processing steps. According to the sensor positions and numbers listed in the configuration files around the site, the wind speed and wind direction measurement units are activated and the output values ​​are recorded by calling internal instructions item by item. The values ​​continuously measured by each sensor within a fixed time interval are sorted in sequence, and the sorted results are compared point by point with the error limit standard obtained by statistics in advance. The error limit standard is derived from the variance statistics of the data obtained by long-term continuous monitoring under windless stable conditions, and an average deviation range is selected as the limit value. When the deviation of any measured data point exceeds the limit value, the data point is marked and discarded, and only the data point set within the limit range is retained. The remaining data points are averaged to obtain the basic wind speed value, and then the wind direction measurement data is processed using the same strategy. The wind direction measurement data converts the sensor readings into quantitative angle values ​​from 0 to 360 degrees through an internal table lookup method, and then the basic wind speed value is integrated with the wind vector quantization value to finally form a set of airflow parameter records at the current moment. The wind speed and wind direction values ​​at the corresponding time are obtained.

[0039] The benefit of the formula is that it directly incorporates the wind field factor into the trajectory point calculation. By introducing the influence of wind speed and wind direction relative to the flight direction of the badminton, the new trajectory point can better reflect the actual flight status under the on-site wind field conditions. The parameter acquisition step is to directly read the trajectory position of the corresponding time point from the previous preliminary trajectory prediction result, and determine it through the image sequence and calibration coordinates ; The parameters are known from the previous stage calculations and are , The wind speed sensors around the site are used to measure multiple times and eliminate the deviation. ; Obtained through the wind direction sensor data mapping table , It has been calculated in the previous step. Determined based on the camera frame rate of 240 frames per second; Calculation process:

[0040] The results show that the updated trajectory points have slight changes compared to the original trajectory points. When the change exceeds the wind field influence reference value (for example, 0.001) obtained through multiple on-site monitoring, it indicates that the wind field factor has a significant impact on the trajectory deviation. If it is less than the reference value, the deviation can be considered slight.

[0041] Through the obtained new trajectory point list, these points are arranged in chronological order, and the spatial distances between adjacent points are compared according to the time label corresponding to each point. When comparing, the deviation limit value contained in the previous statistical record is referred to. This value is obtained by measuring and sorting a large amount of flight data for many consecutive days to obtain the median value, and then a group of data intervals with a smaller distance from the median value are selected as the deviation limit reference. In the comparison, if the distance between any adjacent points exceeds the deviation limit value, the point is marked and eliminated, and the position of the eliminated point is recalculated between the valid points before and after it using the linear interpolation method to obtain the corresponding coordinate value. The interpolation method uses time as the independent variable, divides the interpolation interval into several equal parts, and then performs segmented addition of the coordinate values ​​at the equally divided points to obtain the interpolation result. The interpolated point set is sorted and checked repeatedly to ensure that the distances between all points are within the deviation limit value. Finally, a new trajectory data sequence without abnormal points is obtained, thereby obtaining a processed adjusted trajectory prediction data set.

[0042] The steps to obtain the predicted landing area data are as follows: Based on the adjusted trajectory prediction results, the potential landing point of the badminton in the venue is analyzed to obtain the predicted landing point area coordinates; Based on the predicted coordinates of the landing area, the center and boundary of the landing area are calculated to obtain the center coordinates of the landing area. The calculation formula is: ; in, To predict the number of points in the landing area, is the weight assigned according to the density of the landing points, For the coordinates of each point, is the center coordinate of the weighted landing area; Based on the center coordinates of the landing area, the range of the badminton landing area is determined to form predicted landing area data.

[0043] Specifically, the formula is beneficial in that by performing weighted averaging on multiple coordinate points within the landing area, the center position can more sensitively reflect the density of the landing point distribution. The parameter acquisition step is to determine the number of coordinate points in the predicted landing area obtained in the previous paragraph. For example, the number of points recorded is indivual; The parameter acquisition step is to assign values ​​according to the density of the landing point. First, the number of adjacent points around each point is read from the landing area data, and the density value is obtained by counting the number of adjacent points. Then, the density value is linearly mapped and the result is assigned to the range [0.1,1]. For example, through on-site statistics, it is found that the weight of points with higher density is larger, the weight of points with average density is medium, and the weight of points with lower density is smaller; The parameters are obtained by directly reading the recorded coordinate points of the landing area. These points are all within the range of a standard badminton court (a badminton court is 13.4 meters long and 6.1 meters wide). For example:

[0044] Weight The setting of is based on the number of points around the point. If there are many points around point 1, then , the second point has a higher density , point 3 has medium density , point 4 is slightly lower , point 5 is above average ,These weight values ​​are obtained by collecting the landing point distribution data on site multiple times and performing normalization processing.

[0045] Calculation process: First calculate the weighted sum:

[0046]

[0047]

[0048]

[0049]

[0050] Add the x-weighted values:

[0051] Add the weighted values ​​in the y direction:

[0052]

[0053]

[0054]

[0055] The result shows that the coordinates of the center of the weighted landing point area are close to the dense distribution area of ​​the coordinates of each landing point. When the resulting coordinate value is compared with the previously known boundary line of the venue, it can be found that the center of the current landing point area is located in the middle area of ​​the venue. When the resulting coordinate is greater than 3 meters in the x direction, it means that the landing point is biased towards the middle of the venue. When the y direction is less than 1 meter, it means that the landing point is closer to the area in front of the net.

[0056] Using the weighted center coordinates determined in the previous step, use this coordinate as the center to find the boundary reference value of the corresponding area in the field calibration table. This table is formed by recording the distance distribution between different position points and the center point in multiple measurements. According to the records, the fixed step coordinate increment is determined point by point from the center to the surrounding areas. The point coordinates at each incremental position are read and compared with the established reference matrix. The reference matrix lists distance thresholds from small to large. When the number of matching points in any incremental direction is lower than the number index recorded in the description file, the direction stops extending, and the coordinate step is processed in the same way along other directions. Repeat the above steps for each direction, select all the final coordinates when the extension stops to form a boundary point set, sort this boundary point set and connect adjacent points to obtain a closed area outline around the weighted center coordinates, output the boundary coordinate data of the area in a fixed format, and obtain the predicted landing area data.

[0057] The steps to obtain the precise verification results of the landing point are: Collect the predicted landing area data to obtain the coordinates and boundary information of the badminton landing point; Compare with the site area map to verify whether each landing point is within the site area and obtain the comparison result; Based on the comparison results, the position of each landing point is verified to determine whether it complies with the rules and expectations of the game, and an accurate landing point verification result is generated.

[0058] Specifically, when performing the comparison between the landing point data and the venue area map, the coordinate data of the predicted landing point is first superimposed on the electronic version of the venue map. Verify whether the coordinates of each landing point are within the preset venue valid area. The venue valid area is determined by the official specifications of the venue and is preset in the comparison tool. This process involves a separate verification of each landing point coordinate, and invalidation of coordinates that are outside the preset valid area. All comparison results are recorded in the database to form a comparison result report.

[0059] In the verification step based on the comparison results, each verified valid landing point is tested for compliance with the rules to check whether each landing point meets the requirements of the competition, such as whether the landing point is within the specified scoring area. When performing this test, the official rules of the badminton competition are taken into consideration.

[0060] The present invention provides a badminton landing point prediction system, comprising: An image capture module is used to capture an image sequence of the moment of hitting the badminton, extract the position coordinates of the badminton in each frame, and generate position coordinate data; The dynamic analysis module is used to calculate the displacement and time difference between every two frames of the badminton according to the position coordinate data, deduce the speed and movement direction of the badminton, and obtain the dynamic characteristics of the badminton; The trajectory prediction module is used to predict the flight trajectory of the badminton based on the dynamic characteristics of the badminton by using time series analysis to obtain the position of the badminton at a future time point, and then adjust the flight trajectory in combination with the real-time wind speed data provided by the wind speed and direction sensors around the venue to obtain the adjusted trajectory prediction result; A landing point calculation module is used to calculate the landing point area of ​​the badminton using the adjusted trajectory prediction result and generate predicted landing point area data; The landing point verification module is used to compare the predicted landing point area data with the venue area map to verify whether the landing point of the badminton is accurate and generate an accurate landing point verification result.

[0061] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A badminton landing point prediction method based on computer vision, characterized in that: The following steps are involved: Capturing an image sequence of a badminton hitting moment through a camera deployed on a badminton court, extracting the badminton position coordinates in the image sequence, and generating position coordinate data; calculating the speed and direction of the badminton based on the position coordinate data, and generating a dynamic feature of the badminton; Based on the dynamic characteristics of the badminton, a time series analysis is used to predict the flight trajectory of the badminton, and the expected position of the badminton at several time points in the future is calculated to generate a preliminary trajectory prediction result; according to the preliminary trajectory prediction result and the wind speed data obtained in real time from the wind speed and direction sensors around the venue, the predicted trajectory is adjusted to generate an adjusted trajectory prediction result; Using the adjusted trajectory prediction result, the landing area of ​​the badminton is calculated through a badminton landing point prediction algorithm to generate predicted landing point area data; the predicted landing point area data is compared with a venue area map to verify the landing point and generate a precise landing point verification result.

2. The method for predicting the landing point of a badminton ball based on computer vision according to claim 1, characterized in that: The steps for obtaining the position coordinate data are: The camera deployed on the badminton court is used to capture the image sequence of the moment when the badminton is hit, and the original image data is obtained; Based on the original image data, the displacement of the badminton between two frames in the image sequence is extracted, and the calculation formula is: ； in, is the coordinate of the badminton in the previous frame, is the coordinate of the badminton in the current frame, are the velocity components of the badminton in the x and y directions, is the displacement of the badminton between two frames; Based on the displacement of the badminton between the two frames, position coordinate data is obtained.

3. The method for predicting the landing point of a badminton ball based on computer vision according to claim 1, characterized in that: The steps of acquiring the dynamic features of the badminton are: Based on the position coordinate data, the speed and direction of the badminton are calculated using the following formula: ； and ； in, is the displacement in the x direction, is the displacement in the y direction, is the time interval, For speed, is the moving direction angle; Based on the calculated speed and direction, the dynamic characteristics of the badminton are summarized.

4. The method for predicting the landing point of a badminton ball based on computer vision according to claim 1, characterized in that: The steps for obtaining the preliminary trajectory prediction result are: Based on the dynamic characteristics of the badminton, the movement trend of the badminton is analyzed using time series, the change pattern is evaluated, and a preliminary flight trajectory prediction result is obtained; Based on the preliminary flight trajectory prediction results, using multiple linear regression analysis, the expected position of the badminton at a future time point is calculated to obtain future position prediction data; Based on the future position prediction data, the prediction data at each time point are summarized and compared, and all prediction points are integrated to form a preliminary trajectory prediction result.

5. The method for predicting the landing point of a badminton ball based on computer vision according to claim 1, characterized in that: The steps for obtaining the adjusted trajectory prediction result are: Based on the preliminary trajectory prediction results, collect wind speed and direction sensor data around the badminton court to obtain real-time wind speed and direction data; Based on the real-time wind speed and wind direction data, the badminton flight trajectory is recalculated to obtain a new trajectory point. The calculation formula is: ； in, is the original trajectory point, is the wind speed, is the speed of badminton, is the wind direction angle, is the angle of the badminton's flight direction, is the time interval, is the new trajectory point; Based on the new trajectory point, an adjusted trajectory prediction result is generated.

6. The method for predicting the landing point of a badminton ball based on computer vision according to claim 1, characterized in that: The steps for obtaining the predicted landing area data are as follows: Based on the adjusted trajectory prediction result, analyzing the potential landing point of the badminton in the venue to obtain the predicted landing point area coordinates; Based on the predicted coordinates of the landing area, the center and boundary of the landing area are calculated to obtain the center coordinates of the landing area. The calculation formula is: ； in, To predict the number of points in the landing area, is the weight assigned according to the density of the landing points, For the coordinates of each point, is the center coordinate of the weighted landing area; Based on the center coordinates of the landing point area, the range of the badminton landing point area is determined to form predicted landing point area data.

7. The method for predicting the landing point of a badminton ball based on computer vision according to claim 1, characterized in that: The steps for obtaining the precise verification result of the landing point are as follows: Collect the predicted landing point area data to obtain the coordinates and boundary information of the badminton landing point; Compare with the site area map to verify whether each landing point is within the site area and obtain the comparison result; Based on the comparison results, the position of each landing point is verified to determine whether it complies with the game rules and expectations, and an accurate landing point verification result is generated.

8. The badminton landing point prediction system according to any one of claims 1 to 7, characterized in that: include: An image capture module is used to capture an image sequence of the moment of hitting the badminton, extract the position coordinates of the badminton in each frame, and generate position coordinate data; The dynamic analysis module is used to calculate the displacement and time difference between every two frames of the badminton according to the position coordinate data, deduce the speed and movement direction of the badminton, and obtain the dynamic characteristics of the badminton; The trajectory prediction module is used to predict the flight trajectory of the badminton based on the dynamic characteristics of the badminton by using time series analysis, and obtain the position of the badminton at a future time point. Then, the flight trajectory is adjusted in combination with the real-time wind speed data provided by the wind speed and direction sensors around the venue to obtain the adjusted trajectory prediction result; A landing point calculation module is used to calculate the landing point area of ​​the badminton using the adjusted trajectory prediction result and generate predicted landing point area data; The landing point verification module is used to compare the predicted landing point area data with the venue area map to verify whether the landing point of the badminton is accurate and generate an accurate landing point verification result.

Images