摘要
在半结构化环境下,传统检测算法受到路缘石、坡道、弯道、负障碍等因素影响,算法的检测性能明显下降。因此,提出了一种通过融合道路边界和障碍物检测来提取可通行区域的算法。该算法首先利用约束条件提取局部范围内的特征点及特征直线。其次,通过直线段特征聚类、RANSAC筛选和三次B样条曲线拟合等方式获取完整的道路边界,并采用动态阈值的方式对障碍物进行检索。最后,融合道路边界和障碍物检测结果形成可通行区域。对比实验结果表明,在类似校园的半结构化环境下,该算法能够适应弯道、坡道、负障碍等场景,同时实现更加稳定、准确、快速的可通行区域检测,显著地提升算法性能。
In a semi-structured environment,the detection performance of traditional detection algorithms significantly decline due to factors such as curbs,ramps,curves,and negative obstacles.Therefore,this paper proposes an algorithm,which obtained accessible areas by fusing the road boundaries and the detection of obstacles.This algorithm firstly uses constraint conditions to extract feature points and lines of a particular area.Secondly,complete road boundaries are obtained through feature clustering of straight line segments,RANSAC screening and cubic B-spline curve fitting,with the dynamic threshold used to search for obstacles.Finally,road boundaries and the information of obstacle detection are fused to form the accessible area.Compared with experimental results,in a semi-structured environment similar to campus,this algorithm can be equal to scenes including curves,ramps,negative obstacles,bumps,with achieving more stable,accurate and faster detection of the accessible areas simultaneously,which improves algorithm performance significantly.
作者
庄博
ZHUANG Bo(School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 211094)
出处
《计算机与数字工程》
2022年第11期2435-2442,共8页
Computer & Digital Engineering
关键词
三维点云
半结构化环境
直线提取
曲线拟合
可通行区域检测
3D point cloud
semi-structured environment
straight line extraction
curve fitting
accessible area detection
