거리 영상에서 UV-disparity와 Template matching을 이용한 빠른 사람 검출 방법

Abstract

In this paper, a fast human detection method using UV-disparity and template matching is proposed. The proposed method overcomes a problem of computational complexity presented in conventional approaches. The proposed method improves its performance by adapting size of template with respect to depths of human candidates and finding candidate images. Human candidates images are obtained by using a labeling technique in UV-disparity. Experimental results showed that the proposed method was eight times faster than the conventional approaches using template matching algorithms.