Research On Texture-based Vehicle Image Segmentation And Extraction Algorithm

As the development of the computer vision technology, the application of CCD camera to capture moments of objects can provide more and more high-quality images. The application of image recognition in the field of vehicles becomes more and more extensive. Its main applications are in intelligent transportation and vehicle performance parameters detection. The main examples are in license plate recognition and vehicle type recognition. In the license plate recognition system, followed by the use of license plate location, characters division and character recognition for image processing, where that, in the vehicle type recognition system, the identification process can be broadly divided into the following steps: imagine segmentation, feature extraction. classifier in the department. In the above application can be summed up the core issues and key technologies which consists of the regional division and extraction of the target vehicle and selecting the characteristics of the vehicle. A texture features based segmentation and extraction algorithm for vehicle image is proposed.In this paper, the photographed image and the vehicle lanes vehicle is calculated with the mask. The second process is taking the image of vehicle part away from the picture, then in view of the noise in the picture, providing a reasonable noise filtering method. Then achieve the vehicle image segmentation. In succession analysis the final outline of curve, extract the key line for the calculation of deviation and the measurement of vehicle size. The main contents including:Firstly, the existing methods of image segmentation are surveyed and summarized. A perspective camera model is demonstrated, and the image pretreatment processing operations are introduced, which contain an average Scoreboard-based method to reconstruct lane background, and a median filtering method to improve image quality.Secondly, the algorithm based on texture features of the vehicle image segmentation and extraction is presented, using the key information such as vehicles texture, brightness and color, the optimal threshold is selected through the cycle of iterative method. segmentation and extraction of the vehicle is realized by calculating the texture mask, mask brightness, color mask, after morphological operations and edge detection.Finally, the algorithm proposed in this thesis is realized and tested under the condition of a single vehicle, multi-vehicles, and block (overlap) vehicles. Four factors (correct detection rate, over detection rate, no detection rate, and total area ratio) are measured as the evaluation criterions.The use of texture features on the vehicles and the segmentation and extraction is a bright spot of this article. Vehicle image segmentation and extraction algorithm based on the characteristic of the texture is presented, the simulation is done by one car and overlap using the above algorithm, a platform is provided for the basis of further study.