Improved human detection with occlusion handling from mobile robot

Abstract

In the field of computer vision, surveillance systems are a type of security devoted to the safety of public and property. One of the task of a surveillance system is human detection. This paper presents a human detection system and development of a robust human detection technique using thermaldepth information in an indoor environment from mobile robot. A novel fusion of thermal-depth information (FTDI) is introduced to improve segmentation process, robust in any lighting condition, and to expedite processing. To deal with occlusion handling a new approach is proposed called Occlusion Human Detector (OCHD) with Pre-detector. This detector is used to classify occluded persons using pixel codes that are established from candidate selection process. The quantitative results show that the proposed system performed well with an average accuracy of over 90% for all datasets, and even outperformed state-of-the-art algorithms. The novelty of the work in this paper has presented that the detection method can be enhance the classification of persons and their occlusion. The advantages of the proposed system are that it is computationally inexpensive and performs well even under severe occlusion and poor illumination.