LEG CYCLING TRACKING BY DYNAMIC VISION

This study describes a method of tracking of human body limbs from a m onocular sequence of perspective images. These objects and the associa ted articulations must be modelled. The principle of the method is bas ed on the interpretation of image features as the three-dimensional pe rspective projections points of the object model and an iterative proc ess method to compute the model position in accordance with the analys ed image. This attitude is filtered (Kalman filter) to predict the mod el position relative to the next image of the sequence. The image feat ures are extracted locally according to the computed prediction. Track ing experiments, illustrated in this study by a leg cycling sequence, have been conducted to demonstrate the viability of the approach. (C) 1997 Elsevier Science Ltd.