Proceedings (Feb 2018)
Image Based Stroke-Rate Detection System for Swim Race Analysis
Abstract
Swim race analysis systems often rely on manual digitization of recorded videos to obtain performance related metrics such as stroke-rate, stroke-length or swim velocity. Using image-processing algorithms, a stroke tagging system has been developed that can be used in competitive swimming environments. Test images from video footage of a women’s 200 m medley race recorded at the 2012 Olympic Games, was segmented into regions of interest (ROI) consisting of individual lanes. Analysis of ROI indicated that the red component of the RGB color map corresponded well with the splash generated by the swimmer. Detected red values from the splash were filtered and a sine-fitting function applied; the frequency of which was used to estimate stroke-rate. Results were compared to manually identified parameters and demonstrated excellent agreement for all four disciplines. Future developments will look to improve the accuracy of the identification of swimmer position allowing swim velocity to be calculated.
Keywords