Abstract
In this paper, we propose the new method to track the human head by using the location and color information. Some previous methods, which use multiple cameras, are performed inside two dimensional images. But, we expand it to the tracking in three dimensional voxel space. The voxel space is reconstructed by integrating the depth maps obtained from the multiple stereo cameras. It is necessary to integrate information efficiently to track in real time. Particle filter, which has the robustness for noise, is used to track the human head. The integrated voxel region should be restricted to the minimum necessary for particle filter. Thereby we can obtain images, integrate information and estimate the position of the human head in real time.
