Hand detection based on depth information and color information of the Kinect

Human hand detection refers to the process of detecting human hand from the videos or images. The traditional detection methods based on skin color are easily affected by complex background and light variety, in order to solve this problem, this paper proposes a method combining the depth information with the color information. Firstly, we capture the depth image and RGB color image with 3D camera of Microsoft which is called Kinect, then we select candidate area of the hand by skin color segmentation. Secondly, we segment human hand from the depth image by setting an appropriate depth threshold. Finally, we use the hand extracted from the depth image as a mask for the image of skin color segmentation and then we segment the human hand in the color image. Experimental results show that the hand detection scheme put forward in this paper is easy to implement and has good robustness. Furthermore, it is less affected by complicated illumination and background conditions.