Research On Automobile Driver-Assistance Technology Based On Machine Vision

With the rapid development of the transportation, the traffic safety problems have become more and more serious and the safety of human life and property are threatened by two hundred thousand of the traffic accidents every year. In order to reduce the traffic accidents furthest, scientists put forward the automobile driver-assistance system. Lane departure control is the important content of the auxiliary system. This system can detect whether the car deviates from the lane line in real time and control the direction according to the judgments to reduce the traffic accidents when the driver does not turn on the light of direction. This paper mainly studies that how to recognize the lane information correctly, establish reasonable lane departure model, transmit the deviation information timely and control the direction correctly for automobile driver-assistance technology based on machine vision.Image preprocessing technologies are discussed in order to obtain the lane line information of structured road under the sunny condition. Firstly, extracting and graying the lane line area of image may improve the processing speed and the accuracy of the recognition. Secondly, the principle and methods of image filtering, binaryzation, edge detection are introduced respectively and different methods are tested on the Matlab platform, then median filtering and Sobel edge detection have been chosen. After detecting the edge, the different methods of lane line identification and the Hough transform method are introduced to identify the lane line. Finally, lane line selection method based on gradient is proposed in this paper and relevant parameters of lane line are extracted for the identified multiple lane lines.Compared with the common methods of lane departure, the method based on the rate of lane offset is chosen as the lane departure judgment model and the value of the deviation for different degree of deviation is calculated by theoretical deduction. And then the calculation scheme of the deviation threshold is determined. In order to illustrate the feasibility of lane line identification algorithm, the experiment of lane departure is carried out using the OpenCV platform.The directional control model has been designed on the basis of lane line identification, and the function of different parts is explained. The ECU control module has been simulated using Protues software to explain the rationality of this scheme. The serial port communication has been used to transmit the information of deviation. The EUC module is studied substantively and its control circuit is designed. Then, the fuzzy control method is chosen to control the steering.Finally, an experimental platform is set up to verify the lane departure control by a vehicle running video which is collected by a vehicle recorder. The results show that the recognition rate of this scheme is 0.015s/frame and recognition accuracy rate is 97.9%. The motor rotation direction and rotation angle can change according to the rules under the different deviation in the range of deviation angle of not more than 5°.