A review on vision-based pedestrian detection

Abstract

This paper describes the state-of-the-art in vision-based pedestrian detection, including pedestrian detection methods, training of classifiers and commonly used pedestrian databases. In this paper, we classify the pedestrian detection algorithms into static pedestrian detection and dynamic pedestrian detection, according to the application of motion information. The former is further divided into model-based methods and feature-classifier-based methods, and we mainly focus on the latter ones. Unlike other surveys in pedestrian detection, we also introduce the newest techniques of training classifiers and some general databases of pedestrian detection. Finally, we compare and conclude various pedestrian detection algorithms. Challenges and future trends are also discussed in that section.