Image-based Vehicle Detection And Tracking Of The Doers Of The Word Line To Keep

With the development of technology and society, the per capita consumption of motor vehicles increased rapidly and the traffic safety issues have been the world's attention. The survey shows that many car accidents due to the car far away from the current vehicle line when the drivers are drunk or fatigue. Therefore, line detection and tracking is the important part of intelligent driving assistance system. This paper is the research around the line detection and tracking based on image.First, we do research and experiment on the segmentation based on the brightness estimation and multi-threshold. To deal with the character of camera, we use average filter to estimate the brightness of the road image, then subtracted by the original gray image. The minus image can highlight the line region in the road image. To adapt variety of climatic conditions, lighting conditions and other traffic with complex condition, we use multi-threshold segmentation method to replace a certain threshold which the segmentation result may be inaccurate.Second, we do research and experiment on the regions of interest precession based on estimates of the vanishing point. We calculate the vanishing point to determine the line region of interest (ROI). Then take advantage of the structural characteristics of the line, we propose interference cancellation method based on Run-Length and refine the line with fixed width. Through these steps can eliminate some noises such as:the fence of both sides on highway, the vehicles in front.Third, we do research and achieve the line selection based on the weighted Hough transform and space constraint conditions. We use the weighted Hough transform to extract the candidate lines, and then use the space constraint conditions to select lines. Last, we choose the most suitable pair of lines as detected lines.Finally, the using of Kalman filter to track and keep line. This method can adjust the case according to track the interest regions effectively. Also Kalman filter can shorten extract time and robustness can be further improved.After a large number of experiments show that the method in this paper can get suitable lines, where the method is robust and effective. This research also have some reference on intelligent transportation.