Pedestrian detection: A comparative study using HOG and CoHOG

Abstract

Pedestrian accidents still represent the second largest source of traffic related injuries and fatalities after accidents involving passenger cars. Pedestrian detection is a key problem in computer vision, with several applications that have the potential to positively impact quality of life. In recent years, many pedestrian classification approaches have been proposed. The pedestrian classification consists of two stages: feature extraction and feature classification. Recently several robust feature extracting methods have been proposed in literature like Scale Invariant Feature Transform (SIFT), Histogram of Gradients (HOG), Co-occurrence of Histogram of Gradients (CoHOG). Also several classifiers exists like Hidden Markov Model (HMM), Support Vector Machines (SVM), and Neural Network. In this paper, we examine the two feature extraction method and we use neural network as classifier instead of SVM. An extensive evaluation and comparison of these methods are presented. The advantages and shortcomings of the underlying design mechanisms in these methods are discussed and analyzed through analytical evaluation and empirical evaluation.