Research On Automatic Assessment Technology For Hang Style Long Jump In College Curriculum Teaching

The hang style long jump is a crucial component of athletics instruction in higher education.Currently,there is a lack of standardized quantitative criteria for teaching this skill,leading to subjective evaluation by teachers based on personal experience.This approach lacks objectivity and fails to provide quantitative feedback for long-term tracking and effective evaluation of students’ long jump skills.Consequently,personalized guidance for students is challenging.This thesis addresses the aforementioned issues by proposing an automated evaluation technology for hang style long jump and developing a long jump evaluation system to provide practical tools for quantitative evaluation.The primary objectives of this study are as follows:(1)Measurement of Run-Up Speed in Long Jump: This study proposes a novel method to measure the run-up speed in long jump,utilizing the photoelectric speed sensor principle.The approach-up scene of long jump is modeled using a virtual line speed measurement technique.The YOLOv5 and DeepSORT algorithms are employed for target tracking of the runner.The video frame of the runner passing the virtual line is recorded,and the time for the runner to pass the two virtual lines is calculated by analyzing the frame difference.Finally,the average speed when the runner passes the two virtual lines is computed.To validate the feasibility of the virtual line speed measurement model algorithm,we collected 45 sets of run-up video data with marked points and conducted experiments on videos with different frame rates.(2)Evaluation of Long Jump Aerial Technique Posture: To objectively measure and quantify the posture of long jumpers in the air during long jump,this study proposes an evaluation algorithm based on human body pose estimation.Specifically,we employ the BlazePose pose estimation algorithm to extract joint points of the human body from video frames,reconstruct the topological structure of key points,and establish a set of initial pose standards through comprehensive evaluation by experienced instructors in college-level general training courses.These pose standards serve as the reference standard for automatic scoring in subsequent systems.By using the feature angle of key poses,the tested posture is matched with the reference standard posture,and evaluated accordingly.To ensure the accuracy of the scoring,we fit the scoring function using 30 sets of long jump videos selected by the instructors.Finally,we utilize a posture comparison visualization algorithm to achieve radial transformation,cropping,and scaling of the tested and standard postures,allowing for easy observation and comparison of the gap between the postures.(3)Design and Implementation of Automatic Evaluation System for Long Jump: To meet the actual needs of special training and minor course teaching of long jump in colleges and universities,we design and develop an automatic evaluation system for long jump based on the virtual line speed measurement algorithm and posture evaluation algorithm mentioned above.The system provides convenient quantitative evaluation of long jump for teachers and coaches on the teaching site,with functions including approach speed measurement and aerial technique posture evaluation.The posture evaluation gives results for each key posture,improving the objectivity of performance evaluation and providing posture contrast visualization images and key movement posture feature angle data reports for the reference of teachers and students.This enables timely correction of incorrect actions and improves the possibility of realizing personalized training.