基于混合贝塞尔曲线模型的车道检测算法  

Lane detection algorithm based on mixed Bézier curve model

作　　者：韩浩 王舜燕[1] 

机构地区：[1]武汉理工大学计算机学院,湖北武汉430070

出　　处：《计算机工程与设计》2018年第3期732-737,共6页Computer Engineering and Design

摘　　要：为提高车道检测与重建检测效率,提出基于混合贝塞尔曲线的检测算法,采用基于小区域的扩展搜索方法确定边界区间。扩展搜索通过连续、非连续以及弯曲搜索检测初始位置的边界区间,混合贝塞尔曲线模型用双贝塞尔曲线重建车道曲率变化较大的外边界。该算法包括初始区间的确定、扩展搜索、二值化处理和不同的场景下的车道线拟合,提高了车道重建质量和搜索效率。实验结果表明,该算法应对阴影干扰和光照变化具有鲁棒性,平均每帧处理时间为13ms,在弯道曲率变化较大的车道上表现出91.0%的检测率,在应对事故多发场景时具有较大现实意义。To improve the detection rate of lane detection and reconstruction,a detection algorithm based on mixed Bézier curve was proposed,and an extended search method based on small areas was used to determine the boundary block.The extendedsearch detection method was designed to search boundary blocks including continuous search,discontinuous search and bending search.Mixed Bézier model used two curves to reconstruct the lanes＇ outer boundaries with large curvature variation.The algorithm included initial-blocks＇ determining,extended search,binarization processing and lane boundaries fitting in different scenarios,which improved the quality of lane reconstruction and the search efficiency.The results show that the prosposed algorithm is robust against different shadows and illumination variations.The average processing time per frame is 13 ms.It presentsan 91.0% high-detection rate on curved lanes with large or variable curvatures.In response to accident-prone scenes,it has greater practical significance.

关 键 词：车道检测 车道重建 实时性 扩展搜索 贝塞尔曲线 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]