Research On The Method Of Obtaining Point Cloud Deformation Based On Small Grid And Median

Deformation monitoring is an important means to reflect the health status of the deformable object.Most of the traditional monitoring methods are based on single point measurement,and the number of monitoring points is limited.So it is not sufficient to express the deformation details.As a new way of acquiring spatial data of object,3D laser scanner can obtain high-precision and high-density space coordinates of object in a short time.And it makes up for the lack of traditional monitoring means.However,the same monitoring points cannot obtained because 3D laser scanner is a non-cooperative target measurement.These lead to monitoring difficulties.It is difficult to take both density and accuracy into account for existing solutions,which hinder the application and development of this technology.Aiming at the above problems,this paper focuses on the data processing methods of the deformation monitoring based on point clouds.The main contents are as follows.(1)Introduce the main methods of point cloud preprocessing,including point cloud filtering,point cloud registration and point cloud segmentation.Summarize the existing data processing methods of deformation monitoring based on point clouds and analyze the characteristics of these methods.Finally point out the limitations of these methods.The center of gravity method can obtain high monitoring accuracy,but the number of monitoring blocks is limited by the characteristics of the deformable object.Feature fitting method is suitable for the extraction of artificial targets,however this method consume a lot of manpower and cannot express the deformation details sufficiently because of less number of targets.Model comparison method can obtain high-density monitoring points,however,its calculation accuracy is low because of the impact of modeling error.(2)Aiming at the limitation of existing methods,a method of obtaining deformation monitoring points based on "grid division" is proposed.The core idea is that "the corresponding point group" can be determined by the way of grid division.And "the corresponding point group" can be regarded as high-density monitoring points.According to the shape of the surface after the cloud segmentation,the different method is used.For the quasi-plane,according to the extreme value of the point cloud coordinates,the external rectangular box can be established and can be divided.Then the scanning points that fall into the rectangles are determined.For the irregular surface,the peripheral frame is determined by the boundary line of the point cloud,and the internal grid line can be determined by the direction of the boundary curve.By this way,it can be determined which grid the scanning points belong to.Then grid division is completed.(3)A method of calculating deformation of "corresponding group" based on "median"is proposed.Calculate the shortest distance between the scanned points in the corresponding regions.And then the final deformation value of the corresponding region can be determined by the mean value of the middle part of the distance sorted sort by ascending(decreasing).The deformation value of each sub-region is reduced to the original deformable body,and deformation calculation of the whole deformable body can be completed.The test shows when the observation condition is ideal,the mean square error of the quasi-planar is about± 3mm,and the mean square error of the irregular surface is about ± 6mm.The "corresponding group" obtained by grid division can be regarded as high-density deformation monitoring points.(4)Analyze the influence of noise on the deformation results.The robustness of the proposed method is proved by the test.The influence of grid density on the deformation calculation results is analyzed by the test which shows the smaller cell size,the higher the deformation resolution,and the details of deformation can be more able to be expressed.The recommended number of scanning points in a single cell is given.The method described in this paper can be applied to the field of deformation monitoring,especially suitable for deformable body which is larger and the quasi-smooth.By this method,it is possible to quickly determine the position and the magnitude of the deformation,and provide data support for the deformation analysis and forecasting.