Person Identification Using Top-View Image with Depth Information

In this paper, we describe a novel image-based person identification task. Traditional face-based person identification methods have a low tolerance for occluded situation, such as overlapping of people in an image. We focus on an image from an overhead camera. Using the overhead camera reduces a restriction of the installation location of a camera and solves the problem of occluded images. In addition, we utilize depth information for the identification task. We apply four features to the identification method, (1) estimated body height, (2) estimated body dimensions, (3) estimated body size and (4) depth histogram. Experimental result shows the effectiveness of our method.