In order to reduce the probability of traffic accidents caused by night driving, and provide initiative safety system for vehicles in poor visibility environment, according to the basic requirements of auxiliary driving system, a nighttime pedestrian auxiliary model is designed based on far infrared sensor technology. The model captures the original data source by far infrared sensors. Regions of interest (ROIs) are obtained using gray statistical technique. Multi-scale probability template is constructed, and the data source is matched and detected. The detection rate and missing rate of the model are further improved by multi-frame checksum integrated processing technology. Experiment results show that compared to the common methods, the matching accuracy of probability template of the model is greatly improved. It can be used in the suburbs and urban traffic conditions, and its practicability is good.