The Image Preprocessing Algorithm Research Of Video-based Vehicle Detection And Tracking System

With the continuous development of modern traffic in cities, the number of vehicles grew more and more quickly, it causes urban traffic congestion, traffic accidents occurred frequently. Intelligent transportation system in China in recent years developed quickly. But ITS have very high requirements of real-time processing, so how to improve processing speed of each algorithm in this system of is the key to realize real-time. This topic is generation in this situation.Combining the current related image processing technology and ITS development status, this paper studied some commonly used image preprocessing algorithms based on video of vehicle detection system and put forward the improved algorithm. Design purpose is improving systems processing speed as ensuring accurate identification vehicles and tracking the vehicle.In order to accurately timely for vehicle identification and vehicle tracking, the first step is to confirm the vehicle's exact coordinates in pavement coordinates, so as to get its exact location on the road. Then need to preprocess image before vehicle identification to lay the foundation for vehicle identification. Such as transforming the color images to gray, noise elimination, eliminating the video shaking, eliminating the vehicle shadow, etc.This system mainly divided into two parts, a part of it is the vehicle identification, another part is tracking the vehicle. The part of Vehicle identification, firstly need to transform the color images to gray to accelerate the speed of subsequent pretreatment. Then the image filtering got rid of the noise influence. The video anti dithering eliminate the impact of video jitter. Finally, study the background and subtract the background, then use the threshold value method to eliminate the influence of shadow and extract the ROI. In the vehicle tracking, we use the tracking method based on the characteristics and accord to main characteristics and auxiliary characteristic to track vehicles. To sum up, this paper explored and studied hard in image pretreatment algorithm based on monocular vision's vehicle identification and tracking. The methods have a high application value especially in transforming color images to gray and video anti dithering aspects.