Action recognition for the robotics and manufacturing automation using 3-D binary micro-block difference

Vision-based control systems play an important role in modern robotics systems. An task implementing such a system is developing effective algorithm for recognizing human actions and the working environment design of intuitive gesture commands. This paper proposes action recognition manufacturing automation. The key contributions are (1) fusion multimodal information obtained by depth sensors cameras visible range, (2) modified Gabor-based 3-D binary-based descriptor using micro-block difference, (3) efficient skeleton-based descriptor, (4) combined descriptor. proposed binary difference representation patches from video with complex background several scales orientations leads to informative description scene action. experimental results showed effectiveness on datasets.