ARCHIVES

Accurate race timing and participant identification are critical in competitive sports events. Traditional manual methods often suffer from human errors, delayed result processing, and lack of precision. This paper presents an AI-based automated race winner detection system that identifies the race start using audio signal processing and determines race completion using computer vision techniques. The proposed system utilizes pose detection to detect finish line crossing and Optical Character Recognition (OCR) to extract bib numbers of participants. The system operates in real time at 30 frames per second using mobile camera technology and processes data with minimal latency. Audio detection monitors environmental sound levels to identify the race start signal through decibel threshold analysis. Media Pipe pose estimation identifies key body landmarks such as shoulders and chest to detect when runners cross the finish line boundary. Google ML Kit OCR extracts participant bib numbers from video frames using pattern matching and regex filtering. Race results including bib numbers, timestamps, rank position, and captured images are stored in a local SQLite database. The system enables data export in CSV and JSON formats for further analysis and reporting. Experimental results demonstrate improved accuracy over traditional manual timing methods, reduced human intervention, and real-time result generation capabilities. The system addresses critical limitations in existing race management approaches including timing errors, identification challenges, and processing delays. Although the system demonstrates strong performance in controlled test scenarios, its effectiveness depends on factors such as video quality, lighting conditions, and camera positioning. The project establishes a complete end-to-end framework for automated race monitoring and provides a scalable foundation for future enhancements including cloud integration, multi-camera support, and real-time analytics dashboards.