Research On Registration Of Unorganized Point Cloud

As a new spatial data acquisition technique,3D laser scanning technology has been widely used in a number of areas,such as cultural relics protection, reverse engineering, virtual reality and so on. In order to get the complete data of the measured target,we need to unify the data points collected from different perspective,so that they can be measured in a uniform coordinate system,this process is called point cloud data registration.With the wide application of point cloud registration technology, how to get fast and accurate registration results has become the main target of automatic algorithm for point cloud registration.The main contents are as follows:(1)The initial registration method based on PFH and FPFH descriptors and a sample consensus method is achieved.We complete the initial point cloud registration by using the three main steps of extracting the key points,calculating the descriptors of key points,corresponding point-pairs by using sample consensus algorithm.(2)For the original ICP algorithm,we introduce a voxel grid method to complete the point cloud data streamlining and eliminate the wrong point-pairs by using RANSAC algorithms to achieve the registration of the relatively low overlapping region point cloud data.On this basis, we proposed an improved ICP accurate point cloud registration method.(3)In this paper,we use the above thory to make experiments for three different types of point cloud data.The experimental results show that PFH and FPFH can both complete the initial registration and using the FPFH descriptor can have higher efficiency.We use the improved ICP algorithm to make a further registration based on the initial alignment and achieved higher accuracy.The experiment demonstrates the effectiveness of the algorithm.