A Comparison of Feature Measurements for Kinetic Studies on Human Bodies

Abstract

The paper reports about a performance comparison within a joint project of computer vision, and sport and exercise sciences. The project is directed on the understanding of human motion based on shape features and kinetic studies. Three shape recovery techniques, a traditional technique as used in sport and exercise sciences (manual measurement based on an elliptical zone assumption) and two computer vision techniques (based on a small number of occluding contours, and a new combination of photometric stereo and shape from boundaries), are compared using a mannequin as test object. The computer vision techniques have been designed to go towards dynamic shape recovery (humans in motion). The paper reports about these three techniques and their measurement accuracies.