ROBUST HUMAN DETECTION WITH OCCLUSION HANDLING BY FUSION OF THERMAL AND DEPTH IMAGES FROM MOBILE ROBOT

Abstract

In this paper, a robust surveillance system to enable robots to detect humans in indoor environments is proposed. The proposed method is based on fusing information from thermal and depth images which allows the detection of human even under occlusion. The proposed method consists of three stages; pre-processing, ROI generation and object classification. A new dataset was developed to evaluate the performance of the proposed method. The experimental results show that the proposed method is able to detect multiple humans under occlusions and illumination variations.