Indoor human detection using RGB-D images

Abstract

Recently, RGB-D sensors such as Kinect and Xtion have received considerable attention since they provide depth image that is robust to light variation in the environment. They are mainly used for human computer interaction, surveillance and so on. In this paper, we concentrate on indoor human detection using RGB-D images. Some RGB image based features such as histogram of oriented gradient (HOG) and local binary pattern (LBP) are first briefly introduced. Then, a new depth feature that describes the self-similarity of an image is proposed. Finally, combination of them is utilized to detect the people. This scheme can efficiently describe the humans in the indoor environment. Extensive experiments demonstrate that the proposed scheme can achieve a respective promising detection accuracy of 99.28%, 95.48% and 99.91% on three different collected RGB-D data sets.