Study On Automatic Registration Of Tree Point Cloud Based On TLS

The theme of this paper is concentrate on muti-station single tree point cloud data and muti-station tree scene point cloud data,which are obtained by TLS.The main work of the paper is as follows:1)An automatic registration method for multi-station single-wood point cloud data based on ICP algorithm is proposed.Due to the excessive noise data of the single wood point cloud data,the registration algorithm,the noise point cloud data such as the leaf point cloud,etc.,will cause the registration algorithm to fall into the local optimal solution during the registration process and cannot be accurately registered,resulting in accurate single wood Point cloud data.The automatic registration method based on multi-station single wood point cloud data is divided into three parts.1:According to the characteristics of tree point clouds,extract the feature values of each part,and divide the tree point cloud data according to different feature areas.2: Register different points according to different characteristics,and record rotation matrix and translation matrix at the same time.3: The point cloud is transformed in turn according to the rotation matrix and the translation matrix obtained each time,and the final point cloud registration result is obtained.2)A method for automatic registration of multi-station single-point cloud data based on 4PCS algorithm is proposed.Due to the similarity of the point clouds in the registration process of multi-station single-wood point cloud data,it is impossible to find effective feature corresponding points in the registration,which makes the registration of multi-station tree point clouds more difficult.The multi-station single wood point cloud data automatic registration algorithm based on4 PCS algorithm is divided into the following four parts: First,the data skeleton is generated for the single station single wood point cloud data.Secondly,the tree node analysis is performed on the generated tree skeleton to find the branch point of the tree.Then,according to the tree skeleton nodes,the rough registration of the multi-site single-wood point cloud data is performed.Finally,fine registration is performed on the multi-site single-wood point cloud data after the coarse registration transformation,and the final multi-station single-wood point cloud data registration result is obtained.3)A method for automatic registration of point cloud data for multi-station tree scenes is proposed.Because point cloud data of multi-station tree scenes has a high degree of similarity,when searching for feature corresponding points,it will cause a huge error,resulting in inaccurate registration of multi-station tree scene point cloud data.The registration method of point cloud data based on multi-station tree scene is divided into four steps.1: Extract the tree point cloud from the multi-station tree scene point cloud data,and generate the skeleton of the tree.2: Register the tree skeleton according to the triangle rule,and record the rotation matrix and translation matrix at the same time.3: Apply rotation transformation and translation transformation to the scene point cloud to obtain a new transformation point cloud.4: Perform circular precision registration on the point cloud data of the newly acquired scene.Finally,the registration results of the multi-station tree point cloud scene are obtained.