A monocular visual detection system including twelve detection functions is proposed to assist the road driving for safety. In the lane-mark detection, the lateral inhibition property of human vision system is simulated to improve the detector to satisfy all different weather conditions and to avoid the influence of windshield wiper. The preceding vehicles are detected based on the underneath shadow, left/right borders, and multiple templates of vehicles; then verified by the ratio of lane and vehicle widths, symmetry, and gray-level variance of vehicle regions. The preceding-vehicle distance is estimated based on a single camera with only known focus length and setting height. The pitch and yaw angles of the camera are estimated from the proposed method. The experimental results show that the proposed methods are stable and effective for safety detection in various weather conditions: sunny, misty, dusty, cloudy, rainy day, and night. The average vehicle detected rate is 94.5 %.