Human Detection Based on Fusion of Histograms of Oriented Gradients and Main Partial Features
Chenhui Zhou, Liang Tang, Shengjin Wang, Xiaoqing Ding · 2009
In this paper, a new method for human detection based on Adaboost is proposed: selecting the proper partial features of human which include Haar features and histograms of gradients with new extraction and combining them to form a structure to detect humans. We analyze the robustness of different part detectors of human and gain better features through experiments. And a new method based on histograms of gradients is proposed to reduce the false positives. At last, a whole process framework is constructed for human detection. The results of detection experiments show its validity.