Learning based symmetric features selection for vehicle detection

This paper describes a symmetric features selection strategy based on statistical learning method for detecting vehicles with a single moving camera for autonomous driving. Symmetry is a good class of feature for vehicle detection, but the areas with high symmetry and threshold for segmentation is hard to be decided. Usually, the additional supposition is added artificially, and this will decrease the robustness of algorithms. In this paper, we focus on the problem of symmetric features selection using learning method for autonomous driving environment. Global symmetry and local symmetry are defined and used to construct a cascaded structure with a one-class classifier followed by a two-class classifier. Especially for local symmetric features, different symmetric areas in the rear view image of vehicles are searched through Adaboost based learning, and most useful symmetric features are extracted. The threshold for classification is also found through learning. The effective features selection strategy shows that the integration of global symmetry and local symmetry helps to improve the robustness of algorithms. Experimental results indicate the robustness and real-time performance of the algorithm.