This research project focuses on enhancing the analysis of swimming races by adopting a segmented approach based on race behaviour. Traditional methods of race analysis often overlook variations in performance, race conditions, and individual capabilities. The aim of this study is to develop a novel segmentation method that incorporates these factors to provide more accurate and insightful race analysis. The research is conducted in collaboration with AIMS systems, a company specializing in camera-based race analysis technology.
The study begins with data preparation and segmentation, followed by visualization techniques to analyse race segments effectively. Machine learning algorithms are then applied to identify the most suitable model for evaluating segmented race features. The results are validated and compared against whole race analysis.
Key findings include the development of a segmentation approach that improves race analysis by considering race behaviour and individual capabilities. Visualization techniques, particularly the parallel coordinate plot, seem to be highly effective in analysing race segments for fast races and slow races. The lasso algorithm emerges as the preferred choice for feature analysis in segmented races.
Overall, this research project contributes to the advancement of race analysis techniques in swimming, benefiting both swimmers and coaches in optimizing training and performance strategies.