The Research And Implementation Of Vehicle Logo Detection Method

Along with the progress and development of our country, vehicles are increasingyear by year. At the same time, the intelligent transportation related technologies aredeveloping and improving, in order to be widely used as soon as possible. Theintelligent traffic transportation system, which consists of vehicle license platerecognition system and vehicle logo recognition system, plays an important role in theintelligent transportation. This thesis focuses on the problem of vehicle logo detection.This thesis proposes a novel vehicle logo detection method, which consists of threesteps. First, two images are obtained after using horizontal and vertical direction edgedetection from the preprocessed source image, respectively. Then, a saliency map isgenerated by using visual attention model. And we can get a binary image generatedbased on the result of saliency map. In the end, according to the relevant foregroundinformation obtained from the binary image, we can locate the vehicle logo region.This method mainly has three contributions. First, a novel method is proposed togenerate the object binary image without setting threshold. Second, the vehicle logo canbe detected rapidly without learning. Third, our method is robust in different situationswithout adjusting the parameters. A series of experiments are performed on nearly1000images, which are captured from different real-time situations. Experiment results showthat our method can achieve high vehicle logo detection rate and fast processing time,which are very suitable for real-time applications.