Terrain Points Cloud Matching Using Virtual Grid

With the development of 3D laser scanning technology,the acquisition of large-scale terrain point cloud is more convenient,the scanning speed is faster,and the resolution of the scanner is constantly improved.In the case of surface deformation,it is necessary to compare the point clouds in different periods,such as landslide monitoring and geomorphic deformation;in addition,it is necessary to splice the terrain point clouds acquired by terrestrial scanning point clouds at different stations.When matching the terrain point clouds,ICP algorithm is widely used at present.At home and abroad,scholars have made a lot of improvements on ICP algorithm,but the efficiency and efficiency of these algorithms are not able to fulfill requirements,which need additional auxiliary data and others need pre-processing of the data.Therefore,it is necessary to research the efficient and accurate point cloud matching algorithm for a large-scale terrain.Based on the analysis of the advantages and disadvantages of the classic ICP algorithm and different ICP optimization algorithms,a new matching algorithm using virtual grid is proposed to adapt to the terrain point cloud matching.In this method,the virtual grid is constructed by the point cloud to be matched,and the point cloud to be matched is weighted to get the closest point from the virtual grid node,which reduces the number of the closest point and the optimal transformation point.Effectively it improves the matching efficiency of the point cloud to be matched.Based on the simulation experiments of different types of topographic point clouds,the influence of virtual grid spacing selection on the efficiency and accuracy of virtual grid based matching algorithm is analyzed.With the increase of virtual grid spacing,matching efficiency is improved,but matching accuracy is decreased.When the virtual grid spacing is changing between 1 and 2.5 times average distance of multiple points,the matching efficiency and matching precision can achieve the best balance.Then on this basis,a multi-level virtual grid matching algorithm is proposed,which solves the problem that the matching algorithm using virtual grid is difficult to combine the matching efficiency and the matching accuracy.It has a significant improvement in matching efficiency,which has a good improvement in matching accuracy.In this algorithm,the grid spacing of the virtual grid is transformed step by step during the iteration process.At the beginning of the iteration,due to the large attitude difference between the terrain point clouds,the matching efficiency is guaranteed by the large grid spacing.As the iteration proceeds,the attitude difference of the point clouds becomes smaller,at this time,the virtual grid spacing should be reduced to ensure the matching accuracy.Adopting this strategy,the efficiency and accuracy of terrain point cloud matching can be improved.Two different strategies for grid spacing adjustment in the iterative process are proposed through simulation experiments of different types of topographic point clouds,which changes single step or variable step.The results show that when 1.5 times of the average spacing of multiple points is used as step length,and the effect is better.In this case,compared with the ICP algorithm,the matching efficiency is increased by 95.60%,when the matching accuracy is also improved.The rotation accuracy was improved by 42.86%,when the accuracy of translation is improved by 37.77%.Compared with the ICP algorithm,the multi-level virtual grid based matching algorithm is more suitable for large-scale terrain point cloud matching.