Detection And Tracking Of Human Target Based On Kinect

Target detection and tracking technology play an important role in the field of computer vision. Among them, the detection and tracking technology of the human body target is one of the hot research topics at present. Research results have served in all aspects of the social life deeply. This paper is based on the hardware platform of the Kinect, to test and identify the target of the human body. And analyzing the feature of human body on the basis of the above, realizing the purpose of the tracking of the human body target finally.The Microsoft’s three-dimensional camera Kinect used as the hardware part in this paper. The depth image and color image are acquired through the Kinect. As the most effective detection range of the Kinect is1.2meters to3.5meters. So in the experiment, the targets are processed only in this range. On this basis, the histogram of depth distance is calculated using the depth information of the depth image. And then take advantage of the method of vertical adaptive hierarchical processing to the human body within the scope of the camera. Then get the horizontal projection of the human body in each layer, and extract every human body on segmentation. It plays a big role in solving the occlusions between the human bodies based on the method of the distance segmentation.The method of the index of user is adopted to identify the human body in this paper. If there are multiple human targets in the field of the vision of the camera, then assign a number for each different human body target, and use minimum bounding rectangle to label them. Finally track the human body target using the Kalman filter. The tracking of human body under the condition of occlusion is based on the algorithm of stratification of the depth and the distance, to indentify and mark the obscured human body target. And then realize the purpose of target tracking by using the prediction function of the Kalman filter. In the process of the tracking, the size of the target area and the coordinate of the center point are determined at first. At the same time, present the tracking position and the predicted position in the image for the purpose of testing the accuracy and stability of the Kalman filter tracking. Finally the experimental results show that, the algorithm in this paper can detect and identify human body target correctly under the complicated environment. And there is no influence of illumination variation and occlusion. It can realize the tracking of human body target in the lab environment.