An automated night-time vehicle detection system for driving assistance based on cross-correlation

Recently, real-time vehicle detection become an important task and big challenge in Driver Assistant Systems (DAS), such as collision mitigation and avoidance and dimming of the headlights. However, the most researchers in the field carried out their research in daytime with good light conditions. In this paper, we propose an efficient night-time vehicle detection method in real-time based on image processing. The principle of the proposed method is to detect the vehicle headlamps and rear-lamps using thresholding method and cross-correlation. First, the bright regions (regions of interest) caused by the rear-lamps and headlamps are detected by a white and red color threshold, followed by performing connected component method to label the generated components (potential lamps) by the color threshold, then the potential lamps are regrouped according to their color label and their position and finally, the lamps are filtered by pairing the symmetrical lamps and omitting the other bright regions using the cross-correlation. The experiment results demonstrate that proposed method has good performance in terms of accuracy and robustness and it can meet the requirements in real time. The experiments were done using C++ and OpenCV on Hardware Processor System (HPS) placed in a VEEK-MT2S provided by TERASIC.