The Recognition Of The Vehicle Type And Series Based On GRM Template Matching Algorithm

As the fast development of social economy in our country, people’s living standard has improved conspicuously, meanwhile, the living pace has becoming faster and faster. It is particularly important to have a fast and convenient means of transportation. So it leads to the popularity of automobile in the recent years. At present, people’s demands for car have also been increased. And the price of the car also is acceptably enough. Consequently, the number of cars is growing rapidly on the city roads. The traffic accidents also happen frequently. This will bring the city huge traffic pressures and severe test. In addition, because of the increasing traffic volume, it makes the traffic management work to be heavier and heavier for the traffic managers. For both of these reasons, people hope that there is an Intelligent Transportation Systems to help them finish the traffic management work. It can monitor the traffic conditions in real time.Based on the above background, the thesis will further research the recognition of vehicle type and series by using the relevant theories in the fields of image processing, pattern recognition, computer vision. The thesis realizes the recognition of the vehicle type and series system based on the video images. The main contexts of the thesis are as follows:1. Motion vehicle detection and tracking. The thesis applies the foreground detection algorithm which is the combine of the Gaussian mixture model and self-adaptive background update algorithm to realize the motion vehicle detection.2. The extraction of the vehicle image. In order to accurately extract the vehicle image from the vehicle block which is provided by the foreground detection algorithm. The thesis applies the improved Adaboost algorithm to detect the head of vehicle in the vehicle block. Then we will make use of the position and size of the head to extract the vehicle image.3. The recognition of vehicle type. The thesis applies GRM template match algorithm to recognize the vehicle type. And in terms of the limitations of algorithm application, we will make corresponding improvements on it.4. The recognition of vehicle series. This section includes the coarse location of vehicle logo, the precise location of vehicle logo, the location of vehicle face, the recognition of vehicle logo and the recognition of vehicle face. The way of the coarse location of vehicle logo and the location of vehicle face is a fast and accurately positioning method based on the template of vehicle type. In terms of the precise location of vehicle logo, we firstly analyze the types of vehicles’ air inlet port by using sobel operator. And then we will extract the precise location of the logo by these methods of morphology, horizontal projection, vertical projection and so on. Vehicle logos are recognized by using the method of EHMM in this thesis. The way of the recognition of vehicle face is the same as the recognition of vehicle type.All algorithms of the thesis are realized by using C language based on Opencv 1.0, system interface development is the MFC, development tool is the MS Visual Studio 2005.