基于线性逼近的车道线弯道识别方法

Novel Lane Curve Detection Method Using Linear Approximation

为提高车道线识别算法在大曲率弯道下的识别性能,提出一种基于线性逼近的弯道识别方法.基于车道线先验知识,利用改进的局部逆透视变换和Hough变换对车道线进行初步提取.根据初步提取结果,对未知区域进行循环线性逼近并提取车道线边界点.通过最小二乘法利用B-样条曲线完成车道线拟合.实验证明,该算法对大曲率弯道的车道线识别具有较高的精确性.

In order to improve the performance of lane detection for curve road,a lane detection system was presented with a linear approximation method. Based on the priori knowledge of lane geometry,some initial edge points of the lane were extracted through inverse perspective mapping and Hough transformation. Considering the initial results,a method of linear approximation was applied to search the future lane edge points. Finally,the detected points were used to rebuild the lane according to a B-spline model using least square method. Test of this system shows a promising detection result for curve lanes.