基于消失点引导透视变换的车道线检测算法  

The Lane Line Detection Algorithm Based on the Perspective Transformation Guided by Vanishing Point

作　　者：姚善化[1] 李士杰 王仲根[1] YAO Shanhua;LI Shijie;WANG Zhonggen(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan Anhui 232001,China)

机构地区：[1]安徽理工大学电气与信息工程学院,安徽淮南232001

出　　处：《安徽理工大学学报（自然科学版）》2024年第4期11-19,共9页Journal of Anhui University of Science and Technology:Natural Science

基　　金：国家自然科学基金资助项目(62105004);安徽省高校自然科学研究基金资助项目(KJ2020A0308)。

摘　　要：目的为解决车道线的位置会随着车辆或相机的偏移发生变化而导致车道线检测准确率低和适应性差的问题,提出了一种基于消失点引导透视变换的车道线检测算法。方法首先,采用自适应消失点坐标引导更新透视变换矩阵,将车道图像转换为车道线保存完整的鸟瞰图;其次,将其颜色特征和边缘特征进行融合,得到精准的二值化图像;最后,根据直方图分析定位车道线的基点,采用滑动窗口搜索的方法提取候选的车道线像素,然后对搜索到的车道线像素进行多项式拟合。在不同的道路场景下测试算法的性能,并与其它同类算法进行对比分析。结果仿真结果表明,算法的准确率为94.12%,平均每帧耗时85.35ms,在检测精度和速度方面优于对比的算法。结论该算法能有效解决车道线位置的改变对车道线检测性能的影响,具有更高的准确率和较好的适应性,在阴影遮挡、车道破损、恶劣天气等复杂道路环境的检测下,表现出良好的鲁棒性。Objective A lane line detection algorithm was proposed based on the perspective transformation guided by the vanishing point to solve the problem of low accuracy and poor adaptability of lane line detection due to the fact that the position of the lane line changes with the offset of the vehicle or the camera.Methods Firstly,the adaptive vanishing point coordinates were used to guide the updating of the perspective transform matrix to convert the lane image into a bird's-eye view with complete lane line preservation.Then,its color features and edge features were fused to obtain an accurate binarized image.finally,the base point of the lane line was localized according to the histogram analysis and the candidate lane line pixels were extracted by using the sliding window search method.And then the searched lane line pixels were subjected to a polynomial fitting.The performance of the algorithm was tested in different road scenarios and compared and analyzed with other similar algorithms.Results The simulation results showed that the accuracy of the algorithm was 94.12%and the average time consumed per frame was 85.35ms,better than the compared algorithms in terms of detection accuracy and speed.Conclusion The algorithm is able to eliminate the influence of lane line position change on the lane line detection performance effectively,with higher accuracy and better adaptability,revealing the good robustness under the detection of complex road environments,such as shadow obscuration,lane breakage and bad weather.

关 键 词：车道线检测 自适应消失点 透视变换 特征融合 滑动窗口搜索 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]