Pedestrian Detection for Night View System

A pedestrian detection algorithm, which processes near-infrared images for a night view system, was developed to improve drivers’ awareness of pedestrians in the darkness. The classifier able to discriminate between pedestrians and non-pedestrians in near-infrared images was generated based on the machine-learning framework. By utilizing temporal sequence information, stable and accurate detection performance could be achieved. This research contributed a night view system with a function for pedestrian detection that has been commercialized since 2008. This system allows safer driving at night by alerting drivers to the presence of difficult-to-see pedestrians.