An Effective Skeleton Extraction Method Based on Kinect Depth Image

In order to enable Kinect to achieve skeleton extraction, Microsoft proposed a classifier containing many depth features. To enable the classifier to identify human body, Microsoft input the number of TB-based motion capture data to the cluster system training models. In this paper, we propose a novel human skeleton extraction method based on the depth images extracted by Kinect. Our method does not require complex motion equipment or a large amount of motion data. Firstly, foreground extraction is performed by using the depth information in the depth image to obtain the depth map of the human body area. Then we use the threshold obtained by the algorithm our proposed to segment the body parts with different depth values in the depth map. After segmentation we can obtain the image of the self-occluded part. Next, we obtain the skeleton corresponding to the image of the human body depth map and the self-occlusion part, and finally, we combine the skeletons of these two parts to get the complete skeleton. Experimental results show that our skeleton extraction method can effectively achieve the skeleton extraction of the human body in the natural background.