| Summary: | Current knowledge of tennis player-surface interactions with different court surfaces is limited. The measurement of player step and movement strategy would aid the understanding of tennis player-surface interaction. However, this has not yet been performed: no readily available motion analysis tool is capable of measuring spatio-temporal parameters of gait during match-play tennis. The purpose of this project was to develop, validate and use a motion analysis tool designed to measure player location and foot-surface contacts during match-play tennis. Single camera video footage, obtained from the 2011 Roland Garros Qualifying Tournament, was manually digitised to characterise step and movement strategy during men's and women's forehand groundstrokes. Player movements were consistent with previous notational analyses; however gender differences were highlighted for step frequency. Initial findings were limited by manual analysis, e.g. manual digitising subjectivity and low sample size: an objective and automated system was required. A markerless, view-independent, foot-surface contact identification (FSCi) algorithm was developed. The FSCi algorithm identifies foot-surface contacts in image sequences of gait by quantifying the motion of each foot. The algorithm was validated using standard colour image sequences of walking and running obtained from four unique camera perspectives: output data were compared to three-dimensional motion analysis. The FSCi algorithm identified data for 1243 of 1248 foot-surface contacts; root-mean-square error (RMSE) was 52.2 and 103.4 mm for shod walking and running respectively (all camera perspectives). Findings demonstrated that the FSCi algorithm measured basic, spatio-temporal parameters of walking and running, e.g. step length and step time, without interfering with the activity being observed. Furthermore, analyses were independent of camera view. Video footage obtained from the 2011 ATP World Tour Finals was used to develop a combined player tracking and foot-surface contact identification (PT-FSCi) algorithm. Furthermore, a graphical user interface was developed. The PT-FSCi algorithm was used to analyse twenty match-play tennis rallies: output data were compared to manual digitising. The PT-FSCi algorithm tracked player position and identified data for 832 of 890 foot-surface contacts during match-play tennis. RMSE for player position and foot-surface contacts was 232.9 and 121.9 mm respectively. The calculation of step parameters required manual intervention: this reflected the multi-directional nature of tennis. This represents a limitation to the current algorithm however the segmentation of player movement phases to allow the automatic calculation of step parameters. The analysis of this data indicated that top ranked tennis players can win rallies using movement strategies previously considered to be defensive. Furthermore, step length data indicated that shorter step lengths formed the majority of step strategy. The largest 25% of steps were observed behind the baseline, aligned with deuce and advantage court sidelines. This reflected lunging and turning manoeuvres at lateral extremes of player movement. The single camera system that has resulted from this project will enable the International Tennis Federation to characterise player step and movement strategy during match-play tennis. This will allow a more informed approach to player-surface interaction research. Furthermore, the system has potential to be used for different applications, ranging from sport to surveillance.
|
|---|
