First Sight: A Human Body Outline Labeling System
IEEE Transactions on Pattern Analysis and Machine Intelligence
Model-Based Estimation of 3D Human Motion
IEEE Transactions on Pattern Analysis and Machine Intelligence
Introductory Techniques for 3-D Computer Vision
Introductory Techniques for 3-D Computer Vision
| Hi-index | 0.00 |
This paper presents a new real-time system to acquire pose information of human torso such as body, arm and head. The system does not need any visual marker or device to wear on human body and adopted stereo camera to obtain robust system against for illumination and complex background without position initialization of articulated objects. We present a solution to estimate self-occluded body objects when human model behaves normal action towards the camera. The main idea of the solution is to apply a component labeling techniques on sliced disparity map, and found the arm position when the arm is located in front of basis distance of body and we could also found arm location when the arm is located on the basis distance with Morphological methods. From this method, we can obtain the Torso shape considering self-occlusion. It is simple and fast in comparison with other methods which satisfy real-time performance and accuracy of object tracking at the same time. Furthermore, the method has not boost-trapping problems, it makes the system can work without initialization in the starting. As the result of experiments, we obtained the reliable motion data of articulated object motion under various costume and body type of models.