一种改进的RANSAC算法提取多模型圆弧特征点云

针对传统RANSAC算法迭代次数无上限及只能识别单个模型的局限,提出一种适用于扫描线式点云数据改进的RANSAC算法。对三维激光点云数据进行二维化处理,在RANSAC算法的基础上对局外点进行预剔除,计算过程中对迭代次数进行自适应调整,采用分次识别法实现多模型圆弧点云的提取。实例证明,文中算法能够有效地提取同一场景中的多模型圆弧点云,较传统算法具有明显优势。

An improved algorithm of RANSAC to extract feature point cloud of multi-model arc

Against the limit of traditional RANSAC algorithm which has no caps of iterations and can just identify one model ,it proposes an improved algorithm of RANSAC for scan‐wire point cloud of 3‐D laser scanning .Firstly 3D point cloud is tranformed into 2D point cloud .Based on the algorithm of RANSAC ,the outliers are excluded previously .In the process of calculating ,the iterations are adjusted adaptively .At last the idenitification of multiple arcs is gone through graded recognition method .Experiment show s that this algorithm can extract multiple arcs of point cloud better than the traditional algorithm of RANSAC .