Research On Lane Detection System Based On Car Mobile Platform

The use of machine vision technology for lane detection is an important aspect of theautomotive driver assistance system. In-depth research work done by the key technologiesof lane detection system, focuses on the study and improvement of the vanishing pointbased zoom model. A set of lane detection system which has advantages of good real-time,high accuracy and lower cost for car mobile platforms, has been proposed in this paper.The location of vanishing point is important information in lane feature extraction. Weobtained the vanishing line information by analyzing the relationship of lane image andcamera parameter, and the vanishing point position is calculated by angle vote algorithm,which can greatly improve computational efficiency and increase accuracy.The algorithm of zoom model based on vanishing point chooses the zoom area toexecute comparison operation according to the vanishing point location, and then, extractsthe lane features with some iterative operations. The binary feature map vanishing pointbased of zoom model contains less lane information. Firstly, because of the less laneinformation in the model’s binary feature map, a orientation feature maps with gradientstrength is proposed which can improve the robustness of the feature extraction algorithm.Then, a suitable threshold from the lane feature difference histogram is used here toimprove the efficiency of the lane detection. Finally, in order to solve the problem of a fixednumber of iterations, a terminal condition of adaptive iteration is proposed by analyzing thefeature pixels in comparison operations, which can reduce the calculation time on thepremise of the suitable quality of extracted lane features.In this paper, a linear model has been used to describe the lanes. By analyzingestimation algorithms of lane parameters and combining with the actual situation, a laneparameter estimation algorithm based on tilted projection has been proposed. This methodis robustness of dashed lanes and has better computational efficiency than traditionalHough transform based algorithms.Compared with other common lane detect algorithms implemented on MATLABplatform in extensive experiments, results show that this algorithm can detect the lanes withaccuracy, real-time and robustness, and can be used for car mobile platform.