The Leading Vehicle Distance Measurement Base On Digital Image Processing

Vehicle Location Technology is the key of automatic driving and collision avoidance technology. With the development of computer hardware and software technology, imaging-type distance measuring technology will gradually mature. At present, vision systems used in intelligent vehicles include stereo vision system and monocular vision system. Stereo vision system exists in the image registration problem. Compared with stereo vision, monocular vision with a faster image processing speed and better control of real-time, has good prospects for research and application.A visual range system primarily includes image acquisition, image processing and ranging results output. In this paper, we collect road image by a single camera, then process collected images, finally output ranging results. Image processing section includes:road image preprocessing, line detection of the road, the target vehicle detection and vehicle distance calculation. The main topics are as follows:First, we process the actual acquisition of the road, which includes the image denoising, enhancement, threshold segmentation and edge detection. With the characteristics of road image, in terms of comparative analysis of existing image preprocessing technology and comparative test results by programming, we use logarithmic transformation to improve image contrast, the laplacian operator to enhance image, and the median filter to remove noise. Threshold selection directly impacts on subsequent image processing. After trying to use iteration, segmentation achieves good results. The canny is applied to detect edge and the effect is better.Lane detection:lane detection is to determine the location of vehicles traveling ahead, narrow the scope of treatment and improve the system in real time. Detection methods are generally the least square method, Hough transform and optimal line fitting method. According to the characteristics of the vertical road, using the middle vertical line as the dividing line, we scan the image from bottom to top, respectively from the center to scan both sides. Each line scans a certain number of points and we regard their mean respectively as left or right lane line point. Then using these points to get the best approximation straight line by linear regression, we regard straight lines as a fitting lane line.Target detection and recognition is to find out the object which need to be probed and get the exact location of the vehicle ahead for ranging. Firstly, in the driveway, we use the average pixel gray scale to determine the bottom edge of the obstacle detection for initially identifying the target vehicle position. Secondly, we segment the target objects and background surface by segmentation processing on the road area and mathematical morphology approach is introduced. According to the energy window, we test the authenticity of the target vehicle and confirm the exact location of the vehicle in a small area of forecast.The vehicle distance measurement means using the distance-based model to determine the distance between the main car and the target vehicle. In terms of theory and method of geometric transformation, we establish a coordinate transformation model for monocular camera and create a distance formula based on reverse perspective. We further improve the algorithm to meet the accuracy and real-time requirements in intelligent vehicle control applications. Finally, we obtain ranging results by processing the reality image in terms of test procedures to test the distance effect.