改进的基于累加投影图匹配的点云配准算法

An Improved Point Cloud Registration Algorithm Based on Cumulative Projection Matching

针对大规模点云数据配准拼接中存在的运算耗时长、结果精度低等问题,提出了一种改进的基于累加投影图匹配的点云自动配准算法,该算法首先将三维点云累加投影到二维平面,之后利用尺度不变特征变换算法匹配累加投影图得到匹配点对。在此基础上,根据匹配点对与三维点云数据之间的关系得到三维点云块,最后利用迭代最近点算法匹配三维点云块得到配准结果。实验表明,与Super4PCS算法和基于累加投影图线特征匹配算法相比,本文算法在车载和固定站点云数据配准中都能得到有效的结果,针对大规模点云的处理速度能够提高20倍以上,且能将配准精度提高到0.1 m以内。

The registration and matching for large-scale point cloud data is usually very time consuming and of low accuracy.To solve these problems,this article proposes an improved point cloud registration algorithm based on accumulation projection matching.Firstly,the three-dimensional point cloud data is accumulated and projected to a twodimensional plane,then the matching point pairs are obtained by cumulative projection matching using the scale-invariant feature transform(SIFT)algorithm.The three-dimensional point cloud block is obtained according to the relationships between the matched point pairs and the three-dimensional point cloud data.Finally,the point cloud registration is achieved by matching the three-dimensional point cloud blocks with the Iterative Closet Point(ICP)algorithm.Results show that the proposed algorithm works well for both mobile laser scanned point cloud data and station laser scanning data.Compared withthe algorithm of Super4PCS and line feature matching algorithm based on accumulated projection images,the speed of point cloud registration is 20 times fast,and can achieve the matching accuracy to 0.1 meter.