Vehicle detection using tail light segmentation

This paper presents a method for vehicle detection based on forward looking CCD camera, where vehicle tail light information is employed to generate vehicle candidate. Color segmentation consists of finding pairs of light blobs and removing the isolated points after morphological closing and opening. Among the horizontal light pairs, it determines to define the vehicle candidate. In vehicle candidate verification step, a feature set by Gabor filters using eight direction and five scales is used to train a back propagation neural network (BPNN). In the experiment, this BPNN classifier is used to detect the vehicle. Total 104 images are tested by this algorithm. 87 vehicle images are detected successfully. These results show that our proposed method is effective for vehicle detection in the daytime.