The localization algorithm of human body based on omnidirectional vision

Abstract

In order to meet the needs of moving object localization in real applications, a novel method of human localization based on the omnidirectional vision is introduced in this paper. Firstly, the detection and division of human area is implemented by background subtraction technique and algorithm of connected area labeling in binary image; Then the feet point is obtained according to the imaging characteristics of human body in omnidirectional image. Finally, based on the extracted feet point, the mapping relation between the image plane and the real world plane is obtained through the imaging principle of omnidirectional vision equipment; the position of human body on the ground is obtained. The experiment result shows that this technique is simple and accurate.