Research on Lane Detection Technology Based on OPENCV

Abstract

In view of the huge computing, poor anti-interference ability of traditional detection alogrithm, it does not meet the requirement of the vehicle system, for which this paper proposed a lane detection method based on OpenCV. Preprocessing image in the OpenCV environment, adopting LMedSquare(Least Median Square) idea to select the best subset combined with least squares method to picewise fitting the lane so that it realized automatic identification of lane. This algorithm is suitable for both straight and curve. Simulation shows that this algorithm has well real-time performance, accuracy and robustness. It can meet the requirements of the vehicle system. As the core content of vehicle collision warning system,the lane line detection and recognition widely used in automotive systems.Its accuracy, real-time and robustness are directly related to the safety of the car and the driver.The current research adopts different detection algorithm for straight and curve. This increased the complexity of the algorithm.Hough transform is widely used in the lane line detection, but its computational complexity and real-time performance don't meet the requirements of vehicle system.Least-square method is also commonly used in lane detection, but its sensitivity to noise and poor anti-interference abilityaffected the accuracy of lane line identification(1). To solve problemsabove, this paper put forward an idea which based on LMedSquare to select the best subset, combined with the least square method to piecewise fitting algorithm.This method can eliminate the unnecessary noise in the search process for the lane line feature points so that improving the system's anti-noise ability. The algorithm applies to both straight and curve and reduces the complexity of the algorithm. At the same time,it meets the real-time requirement of the vehicle system.