Research Of Lane Detection System Based On Machine Vision

With the development of the world economy and the improvement of people’s living level, people’s requirements to quality of life are also getting higher and higher. With the constant improvement of the people of car ownership, the increasing of cars on the road gradually increases the incidence of traffic accidents. Therefore, in order to allow the driver to safer driving on the road, research on prevention initiative security technology is particularly important. The lane detection based on machine vision is becoming a hot spot of current research vehicle auxiliary driving system research topic, and it is also an important component part of the study on "intelligent transportation system (ITS).As a multi-disciplinary cross-technology, Machine vision technology simulates human visual function by computer, which not only play acts as human eyes but also instead of the function of the human brain. It uses the image sensor to achieve real-time monitoring of the front of the vehicle on the road environment. It can detect obstacles on the road in detecting lane line at the same time. It also can realize traffic monitoring, and lane detection realized by using machine vision is the most important of the mature technology. It is low cost, reliable performance, easy installation, simple and popular, it can greatly reduce the possibility of accidents, what’s more, it has certain practical significance to the study of lane detection system.Based on machine vision Lane Detection System, the system is an indispensable part of Driver Assistance based on safety the premise, but also for intelligent traffic, mainly for the protection of the vehicle correctly in the driveway, in order to reduce vehicle lane departure caused by traffic accidents. This paper is based on the CCD camera as the image acquisition sensor, based on machine vision algorithm as the image processing algorithm, the embedded DSP as the hardware platform of the lane detection system. This system is not only suitable for the straight road, but also applicable to the curve of the road and the method used in this paper is validated by the test.This topic is the research cooperation with Chery Automobile Research Institute of our project, provides the feasibility and practicality for the system research. This paper analyzes the methods of camera calibration for CCD camera, and then, develops a calibration method for high precision of the camera calibration, establishes the relationship between images and objects space, and realizes the restoration and reconstruction of three-dimensional images. Due to the road environment variable, the road image is easily affected by external environment, the image in the process of gathering and transmission can cause a drop in the quality of image, therefore, we want to make a series of processing on the image, here we adopt an efficient method of median filter, reduce the time complexity of the algorithm, improve the timeliness. Then, in the part of lane detection algorithm based on the proposed method, this paper put forward to a new region of interest, increase the processing speed of image, image processing algorithm to reduce the error rate.The lane model of linear parabolic model, and through the analysis of the traditional Hough transform, we put forward a kind of Hough transform algorithm based on gradient. In the aspect of track, according to the analysis to the existing algorithms and combined with the lane model, we use Kalman filtering on the lane line tracking, using Kalman filtering update lane model.Finally, this article will be transplanted into the DSP platform, the whole system has carried on the experiment; the experimental result demonstrate the feasibility and effectiveness of this system, and meets the real-time requirements.