Towards model-based recognition of human movements in image sequences

摘要: Abstract The interpretation of the movements articulated bodies in image sequences is one most challenging problems computer vision. In this contribution, we introduce a model-based approach for recognition pedestrians. We represent human body by 3D-model consisting cylinders, whereas modelling movement walking use data from medical motion studies. estimation model parameters consecutive images done applying Kalman filter. Experimental results are shown synthetic as well real data.