Pedestrian detection in complex scene using full binary tree classifiers based on locally assembled Binary Haar-like features

Abstract

Under complex scene urban environment, in order to detect pedestrians efficiently and accurately, we propose a high real-time and robust performance pedestrian detection method based on machine vision in this paper. Firstly, a new feature called Locally Assembled Binary Haar-like (LABH) is selected as the feature vector. In this novel feature, Haar features keep only the ordinal relationship named by binary Haar feature, then, brings in similar idea of Local Binary Patter (LBP), assemble several neighboring binary Haar feature to improve the ability of illumination invariant and discriminating power. Furthermore, a full binary tree structure is presented to build on an efficient classifier, which has advantages of both series connection structure and parallel connection structure and brings in a principle of “Early-rejection”, could improve system's real-time performance. The experiment carried out on videos from INRIA dataset, MIT dataset and Daimler dataset illustrates that the proposed method is real-time and feasible enough for pedestrian detection in intelligent vehicle environment.