Research On The Techniques Of Coarse Registration Of Point Clouds

With the maturity of three-dimensional scanning technology,and large-scale application of laser scanners,structured light scanners and other related equipment,the 3D modeling,cultural relics and other work has been greatly facilitated.It is very convenient,simple and efficient to use the 3D scanner to obtain the geometric information of the object.However,it is limited by the scanning range of the scanner,the size of the object,the geometry of the object and so on.We can not get complete information through only one scan.To obtain the complete geometric information of the object,we must scan several times,and then register different views of the data.There are many algorithms for data registration.The 4PCS algorithm is one of the best algorithms in the field of coarse cloud registration.This paper presents a new algorithm to solve the registration problem.First,we propose a new way to choose a 4-point basis.The boundary selection is used to assist the selection of the base,which improves the accuracy and validity of the selection,reduces the search space of the base and improves the search efficiency.After the collection of the bases,we remove the invalid bases by using the limitations of the features,thereby reducing the validation time of the algorithm and improving the efficiency.We experimented with the relevant datasets,and experiments show that our method is more efficient than the 4PCS method.In addition,this paper also proposed an improved registration scheme based on RANSAC method.This method improves registration accuracy by removing mismatched pairs of points.Firstly,the proposed method uses the feature matching method to find the corresponding set of feature points.This paper compares the advantages and disadvantages of various methods of extracting key points.By selecting the appropriate algorithm to extract the key points of the source cloud and the target cloud,we can get the set of corresponding points using the method of feature matching.The original set of corresponding points may contain many dis-matching point pairs.By removing dis-matching point pairs,we can improve the accuracy of registration.