Athlete Behavior Recognition Technology Based on Siamese-RPN Tracker Model

Area of Science:

• Computer Science
• Artificial Intelligence
• Sports Analytics

Background:

• Deep learning algorithms are increasingly applied in Unmanned Aerial Vehicle (UAV) driving, visual recognition, and target tracking.
• In sports, deep learning facilitates athlete trajectory and behavior capture through target tracking and recognition.

Purpose of the Study:

• To develop an advanced algorithm for tracking and recognizing athletes' behavior in sports.
• To improve the accuracy and robustness of behavior recognition models by minimizing environmental interference.

Main Methods:

• Proposed a Siamese Regional Proposal Network (RPN) algorithm combined with an adaptive updating network for athlete behavior tracking.
• Utilized athlete behavior as target candidate boxes during model training to reduce environmental interference.
• Established a simulation model for behavior recognition.

Main Results:

• The Siamese-RPN algorithm effectively reduces background and environmental interference in tracking athlete behavior trajectories.
• Demonstrated improved accuracy and overall performance of the behavior recognition model.
• The algorithm successfully ignores background interference elements in behavior images.

Conclusions:

• The Siamese-RPN algorithm offers superior performance for sports behavior recognition compared to traditional twin network methods.
• It enables offline operations and effectively distinguishes environmental interference factors.
• The algorithm's ability to quickly capture athlete behavior characteristic points highlights its practical application potential.