Decision Tree Classifier Based Pedestrian Detection for Autonomous Land Vehicle Development

Pedestrian detection and accident avoidance system plays very important role in development of autonomous vehicle. A system which can detect pedestrian accurately and take action accordingly can avoid happening the misfortune. This paper proposed pedestrian detection system based on combined information from the decomposition of body part. Three different cascade classifiers trained with different features of body part. Aggregated channel features consumed to train the full body of pedestrian detector while Haar like features utilized to train upper body and face detector. Decision tree algorithm evaluated the output of different body part detector and formulated final decision about the existence of pedestrian accordingly. Body part decomposition and different features utilization by the system makes system more accurate detector under different hazardous condition. Achieved results show that proposed pedestrian detection system works effectively.