Multiple Probabilistic Templates Based Pedestrian Detection in Night Driving with a Normal Camera
Mei Hu · 2006
Pedestrian detection is particularly challenging, comparing with other targets in the domain of object detection, especially for night driving just with a normal camera. In this paper we combine two probabilistic templates based classifiers for elaborate pedestrian detection: the binary probabilistic template based classifier (BPTC) as the first layer to reject most of non-pedestrians by the features of binary image; the gray probabilistic template based classifier (GPTC) as the second layer to make the final classification by the gray probability, which is the contribution of this paper. Experiments show that our approach performs well most of the time, and the system can achieve realtime detection.