| dc.description.abstract | Developing real-time automotive driver assistance systems to alert drivers for possible collision with other vehicles has attracted lots of attention lately. In this study, we use cameras mounted on a vehicle to capture road scenes for forward lane detection, preceding vehicle detection, and blind-spot visual detection.
The forward visual system consists of ten functions: lane detection, multiple lane estimation, classification of solid/dashed lane marks, direction estimation, vehicle lateral offset estimation, lane departure warning, preceding vehicles detection, distance of preceding vehicle estimation, brake light detection, and turn signal detection.
The blind-spot system contains four modules: near lane mark detection, classification of solid/dashed lane marks, side vehicle detection, and distance estimation.
In the proposed system, the lane marks are detected by searching the optimal parameters of a defined lane model on the images. Preceding vehicle is detected by underneath shadow and left/right borders; then use ratio of the road width and vehicle width, symmetry, and variance to verify vehicles. The proposed vehicle detector uses template matching, symmetry, and appearance times to detect brake light, and uses template matching, position check, cluster, and frequency to detect turn signals. Side vehicle is detected by underneath shadow and left/right borders.
In the experiments, the proposed detectors were evaluated on several different weather conditions such as sunny day, misty day, dusky day, cloudy day, and dark night. From the experiment results, we find that the proposed approach can stably detect or track the lanes and vehicles in real time. | en_US |