Computer Vision Algorithms for Human Movement Analysis in Sports Activities

The project team focuses on studying and improving computer vision algorithms for analyzing human motor activity, with special attention to enhancing triangulation methods and classifying movements in sports exercises.

Team Members: Project Lead: KBTU Master’s student Farhad Amanbayev KBTU Undergraduate Students: Rishat Zhanat, Guldanа Rishadkyzy, Mariya Mamedova

Scientific Supervisor: Professor A.Zh. Akzhalova, KBTU SITE (Director of the SDG Projects Department, Academy of Corporate Education)

Main Objectives

• study existing triangulation methods (polynomial and classical approaches)

• improve triangulation algorithms for two-camera setups

• explore methods for keypoint extraction (MediaPipe, OpenPose, etc.)

• test classification algorithms on sports movements (running, pull-ups, squats, etc.)

• perform comparative accuracy evaluation

• prepare a final report/scientific article

Technologies and Tools Used

• Python

• OpenCV, MediaPipe, OpenCap

• NumPy, Pandas

• Matplotlib

• CSV/JSON datasets

• Human Activity Recognition (HAR)

Expected Results

• implemented and improved triangulation algorithms

• experimental accuracy results

• scientific publication

• prototype system for athlete movement analysis