3D OBJECT TRACKING AND MOTION SHAPE RECOGNITION

Annotation. Object tracking and its movement classification is interdisciplinary topic which can be used in many domains like robotics, video surveillance, multimedia systems and etc. Even though it has been addressed in many works it is still challenging task. Many of the existing approaches have such disadvantages like illumination dependency, calibration problems or working only in 2D space. In this paper we propose 3D tracking which does not have above disadvantages because of depth camera that we use here. We propose to identify objects by their HSV values of their colors. Then we calculate coordinates of the center of our interested object by the principle of the center of mass and after that transform them into real world 3D coordinates. Finally we apply Hidden Markov Model to recognize motion shape that was performed by our object. Experiment results demonstrate the efficiency of proposed method.