Vehicle Type Automatic Recognition System

Vehicle recognition system mainly consists of two key technologies: recognition of vehicle license plate (VLP) and recognition of vehicle type. It not only finds wide application in ITS, but also it's a hot point of research in computer vision, image processing and pattern recognition. So its related technology is attended prevalently. In this paper, we pay attention to the localization and recognition of vehicle shape and logo. The methods to deal with above problems are proposed in this paper and testified in the experiment.The paper includes the following parts:1. Introduction of image processing and object recognition: Image pre-processing is very important because it'll determine the result of the following recognitions. Object recognition algorithms are key problems in this recognition system. In this section, we'll introduce the main technology of image processing that used in this system at first, then the prevalent object recognition measures will be presented.2. Detection and recognition of vehicle shape: Recognition of vehicle type according to shape, size is important technical problems in recognition system as well. In this section, we eliminate the background by difference algorithm between the background image and vehicle image. Then we can detect the edge of the vehicle and base on the size to judge the type of vehicle.3. VLP localization algorithm: Vehicle logo localization is based on the location of VLP in this system, which will directly affect further process and is a key step to improve the precision of vehicle logo localization. An edge project method is proposed in the paper and the experimental results show that the measure can meets the need of real-time system.4. Vehicle logo localization and recognition: In view of most vehicle logos are make up of the basic geometry shape such as round or ellipse, so in this section we can find the region including vehicle logo by the geometry location relation between the VLP and the logo, then use the predefined geometry shape to locate the logo by template matching algorithm which based Hausdorff distance. For some complex shape logos, we can use the logo template to locate the logo directly by template matching algorithm. This algorithm can not only get the accurate position of vehicle logo, but also we can use the result of template match that calculate relative index using vehicle logo standard template, and use matching value to recognize vehicle logo for the first time. Based on the first recognition, the global feature of vehicle logo is extracted by edge direction algorithm to match feature and recognize vehicle logo.5. Research of other identification mark: There are a lot of research about the vehicle license plate and logo, whereas no body does enough work about the detail information such as the different kinds of vehicle belong to the same brand and the identification of exhaust. The last parts of this paper do some research about the above aspects.