A Novel Pedestrian Detection Method Based on 2D Histogram Template and HOG Describer in Infrared Images

摘要

Pedestrian detection is a challenging problem in many applications of computer vision. Because visible image based methods have difficulty in extracting moving objects from complex backgrounds, more and more researchers turn to infrared images. In this paper, a novel method for detecting pedestrian in infrared video images was proposed. This method includes two steps: the first step is to describe the pixels with large intensity in images by the univariate Gaussian model and segment the ROIs of candidate objects, the second is to detect human objects by utilizing the coarse classifier based on 2D histogram template and the fine classifier based on HOG describer. Experimental results show that the proposed approach has a high detection rate as well as a low false alarm rate for real-time processing.