Study On Methods For License Plate Detection And Car Type Classification

With the rapid development of the economy, the number of vehicle is growing. The research on intelligent transportation systems (ITS) behaves important. The vehicle recognition system in ITS field has a wide range of applications. Image processing and computer vision technology plays an important role in the vehicle recognition. This paper describes the key technologies of vehicle recognition from the perspective of computer vision. In this paper, the license plate, vehicle logo and car type recognition will be discussed deeply.For license plate, this paper proposes a fast and accuracy method of license plate location and its orientation correction based on the information of the characters. Find the specific location of license plates through the number of connected components and their position. Then, make orientation correction according to the center of connected components. This step contains two points, horizontal and vertical correction. Horizontal tilt angle can be got by the line through the center of the connected components. Vertical correction can be done by projection and pixels translation. The experimental results show that the proposed method can performance well even in the complicated conditions. The results of license plate location and the orientation correction are very good. In addition, for very non-uniform illumination conditions (shadow and exposure), the AdaBoost algorithm is used in this paper to locate the license plate.For vehicle logo, in order to locate the car logo in the various complex scenes, firstly, the AdaBoost algorithm is used to detect the car face, and then coarse location of the logo is completed through the car face. Next, a fine logo location is done based on edges and contours. Finally, for the logo Recognition, SURF and LBP features are respectively used for recognition, and their performances are compared and discussed. The experimental results show that the proposed method is practical and effective, and also be robust to scale and rotation changes.For car type recognition, vehicle shape and size are not the focus. The recognition of vehicle type is discussed, mainly using the local features. SURF and LBP features are respectively used for recognition, and their performances are compared. Finally, classifiers with two stages are proposed.In this paper, the key technologies of license plate location and type recognition are discussed and a lot of useful experiments are showed.