A two-layer night-time vehicle detector

We present a two-layer night time vehicle detector. At the first layer, headlight detection [ref] is conducted to allocate areas (eg, bounding box) where are the possible pairs of the headlights in the image, the Haar feature based Adaboost framework are then applied to decide the vehicle front at night time. This approach has achieved significant performance for vehicle detection at night time. Our results showed that the proposed algorithm can reach over 90% of detection rate at 1.5% false positive rate. Without any code optimization, it also performs at a faster speed compared to Haar feature based Adaboost approachWeihong Wang, Chunhua Shen, Jian Zhang and Sakrapee Paisitkriangkra