Fast Human Detection by Boosting Histograms of Oriented Gradients

In this paper, a novel real-time human detection system based on Viola´s face detection framework and Histograms of Oriented Gradients (HOG) features is presented. Each bin of the histogram is treated as a feature and used as the basic building element of the cascade classifier. The system keeps both the discriminative power of HOG features for human detection and the real-time property of Viola´s face detection framework. Experiments on Daimler Chrysler pedestrian benchmark data set and INRIA human database demonstrate that this framework is more powerful than Viola´s object detection framework on human detection.