Abstract: Considering the complementarities of data obtained by radar and computer vision, a method of preceding vehicle detection using the image and the radar data was proposed. The improved DBSCAN was proposed to process the laser data according to their characteristics. During the vehicle detection, the vehicle hypothesis was generated with the fusion of laser data and image. After vehicle hypothesis formation, then, features were fuzzed and the vehicle detection was realized according to the confidence obtained by fuzzy inference system. The results indicate that the algorithm can detect vehicle without constricts of lane lines and it has good adaptive ability and anti-jamming ability. It can detect the proceeding vehicles accurately.