Automatic Pedestrian Recognition Using Real-time Motion Analysis

This paper presents a real-time vision system for automatic pedestrian recognition using image sequences analysis. We first introduce a robust motion detector based on a specific background structural model. The motion detection process exploits the structural difference between the model and features extracted from the current scene. The pedestrian recognition process analyses the signal provided by the motion detector in order to separate pedestrians from other moving objects (vehicles, buses, ...). The detection signal processing consists in the extraction of several kinematics and pattern features of each classes of objects. The percentage of good pedestrian discrimination, tested with real traffic image sequences under several illumination conditions, is about 85%.