基于ASIFT图像特征的大视角点云配准方法
On the Large View Angle Point Cloud Registration Method Based on ASIFT Image Features
Received:July 13, 2017  Revised:August 28, 2017
DOI:10.7520/1001-4888-17-146
中文关键词:  点云配准  图像特征  仿射尺度不变特征变换(ASIFT)算法  改进的随机抽样一致(RANSAC)算法
英文关键词:point cloud registration  image features  affine scale invariant feature transform (ASIFT) algorithm  improved random sample consensus (RANSAC) algorithm
基金项目:国家自然科学基金(11332005), 上海市自然科学基金(17ZR1410700)
Author NameAffiliation
SHEN Ze-wan Department of Mechanics, Shanghai University, Shanghai 200444, China 
LI Kai* 1.Department of Mechanics, Shanghai University, Shanghai 200444, China 2.Shanghai Key Laboratory of Mechanics in Energy Engineering, Shanghai 200072, China 
Hits: 1413
Download times: 379
中文摘要:
      点云配准是物体三维轮廓获取的重要环节。基于图像特征的点云配准方法能较快地实现这一目标。然而,这种方法对于大视角点云的配准并不理想。针对这一问题,本文提出一种基于ASIFT图像特征的大视角点云配准方法。首先,用ASIFT算法提取大视角图像间的二维匹配点对;而后,利用图像像素和点云间的对应关系得到大视角点云间的三维匹配点对;最后,提出一种改进的RANSAC算法实现点云配准。实验结果表明,该方法可成功实现大视角点云配准。
英文摘要:
      Point cloud registration is an important link for three-dimensional (3-D) profile measurement. Point cloud registration algorithm based on image features can achieve this goal rapidly. However, this method is not ideal for large-viewing-angle point cloud registration. To solve this problem, a large-viewing-angle point cloud registration algorithm based on ASIFT image features is proposed in this paper. Firstly, two-dimensional (2-D) matching point pairs between large-viewing-angle images are picked up by using ASIFT algorithm. Then, three-dimensional (3-D) matching point pairs between large-viewing-angle point clouds are obtained by using the corresponding relationship between image pixels and point clouds. Finally, an improved RANSAC algorithm is proposed in this paper to realize point cloud registration. Experimental results show that this proposed method can successfully realize large-viewing-angle point clouds registration.
View Full Text  Download reader
Close