Real-time Pedestrian Detection Using a Boosted Multi-layer Classifier

Techniques for detecting pedestrian in still images have attached considerable research interests due to its wide applications such as video surveillance and intelligent transportation systems. In this paper, we propose a novel simpler pedestrian detector using state-of-the-art locally extracted features, namely, covariance features. Covariance features were originally proposed in [1, 2]. Unlike the work in [2], where the feature selection and weak classifier training are performed on the Riemannian manifold, we select features and train weak classifiers in the Euclidean space for faster computation. To this end, AdaBoost with weighted Fisher linear discriminant analysis based weak classifiers are adopted. Multiple layer boosting with heterogeneous features is constructed to exploit the efficiency of the Haarlike feature and the discriminative power of the covariance feature simultaneously. Extensive experiments show that by combining the Haar-like and covariance features, we speed up the original covariance feature detector [2] by up to an order of magnitude in processing time without compromising the detection performance. For the first time, the proposed work enables covariance feature based pedestrian detection to work real-time