Research On Refined Modeling Method Based On TLS Technology And Close-range Photogrammetry Technology

At present,it becomes easier to acquire point cloud data,such as the development and application of Kinect depth camera,3D laser scanning technology and digital photogrammetry technology.In order to solve the problem that how to convert 3D point cloud data into three-dimensional expressive entity with certain functions and structures quickly and accurately,Professor Yang Bisheng of Wuhan University put forward the concept of “generalized point cloud”.Each kind of sensor has its own advantages and disadvantages when it comes to obtaining three-dimensional spatial information of the target surface.After multi-sensors,the multi-platform spatial data formed by the multi-source sensor point cloud data is the generalized point cloud.The technology of three-dimensional laser scanning can obtain the three-dimensional coordinates of the target surface with high accuracy,but it is easy to lose the data information of the top for the larger target objects.If the top features of the target object are rich,the model distortion will be caused.The digital photogrammetry technology can obtain spatial information of the target surface in easy,quick ways and multiple perspectives.However,the downside is that the quality of the point cloud data is not easy to control,and the quality of the point cloud data is directly proportional to the time complexity when processing the generated point cloud data.Therefore,guided by the concept of the generalized point cloud,it is worthwhile to study the problem that how to apply the three-dimensional laser scanning technology and digital photogrammetry to the HighPolygon Modeling of large objects.This paper first introduces the basic working principle of terrestrial 3D laser scanner,analyzes its main sources of error and gives the general working flow of the laser scanner.Besides,using the gridding modeling method and the contour feature geometry modeling method respectively,real three-dimensional reconstruction of the gates of school buildings and regular buildings with irregular large-scale structures was realized.Secondly,the principle of coordinate registration is introduced.Five common coordinate registration methods of Riegl vz1000 are experimentally analyzed.Experiments show that in relative registration,registration accuracy based on feature points with the same name is higher than the non-target characteristic point registration,but a large number of targets need to be deployed.In the absolute registration,the accuracy of registration based on the control point of the same name is higher than that of the two rear-view registration methods.Compared with the detection point,the error in the space point is 0.0028 meters,and the coordinate conversion accuracy is higher.Finally,it uses special software of drone measurement,the Pix4 Dmapper,to acquire the surface spatial data of the target and combines with the laser scan data to generate a three-dimensional point model with integrity and high quality.Besides,mesh reconstruction is also realized.The hole filling effect at the top of the object is closest to the target entity compared to Multi-hole algorithms.