Research On Road Vehicle Logo Information Detection

Along with China's reform and open policy and the accelerated urbanization process,the number of cars has increased rapidly,following the unreasonable urban planning,which have brought huge challenge for the traffic management.Nowadays,the intelligent transportation system has gradually become the most powerful traffic management technology.As one of the key technologies in the field of intelligent transportation,the recognition technology of vehicle information has very high practical value.Road traffic information detection includes LPR(License Plate Recognition),vehicle classification,vehicle-logo recognition,etc.Current researches focus more on license plate recognition and vehicle classification,while the research objective in this paper is vehicle-logo detection and recognition.In most realistic scenarios,the surveillance are subject to pedestrians,trees,buildings,billboards and others,so it is difficult to detect vehicles information.To solve these problems,this paper has carried out the following studies:(1)Select actual scenes from road monitoring video to get image preprocessing with image processing technologies such as graying,the image smoothing,Sobel edge detection,binarization,etc.(2)Based on the license plate character of rich edge,we apply the edge image to realize the coarse position of license plate.Through the structure characteristics of the license plate,the unsatisfied candidate area could be removed.Finally,the license plate can be accurately located based on the texture features of license plate characters and vertical projection.(3)By using prior knowledge of the relation between license plate and vehicle-logo,as well as the license plate positioning,we could locate the coarse position of logo.Then through a series of preprocessing,the inteference of noise could be removed.We could conduct a comprehensive analysis of the geometrical characteristics of the logo and texture characteristics,combined with the features that the central axis and the badge logo are symmetry,to achieve accurate positioning.(4)Set up a normalized processing logo library,and use support vector machine classifier to identify logos.Experimental results show that the designed algorithm has a good practicability for detecting the needed vehicle information,and this algorithm is simple andeffective,so it serves a great significant role in the accurate detection of vehicle information in the intelligent transportation system...