Abstract
In this paper propose a robust human detection system based on combined with Higher-order Local Auto-Correlation (HLAC) features and Histograms of Oriented Gradients (HOG) features. HLAC features are invariant to shift. HOG features are robust to change illumination. To combine these two features, we use the co-occurrence of multiple features which contained HLAC and HOG. The co-occurrence features are generated by the Real AdaBoost. We conduct human detect experiments using INRIA database. Beside we confirm influence of changing the number of training images. The experiment results show that the propose method is more effective than the conventional method, and suggest that the there are the optimal numbers of training images.
