Research On Spatial Registration For Three-Dimensional Laser Scanning Data

Construction of virtual geographical environment depends on multiple technologies,of which the development of data acquiring technology provides a platform of the data support.Among the existing data acquiring technologies,3-D laser scanning is so competent to the real world's digitization that it makes other technologies in the shade.In order to reconstruct 3-D model of object's surface,the first step is to scan it in all directions by 3-D laser scanner.However,it's impossible to obtain all the point clouds of the complex object in one time due to the restrictions of observing environment,instruments and the object's shape.Therefore,we should firstly scan the object in all directions by laser scanner,and then match the multi-viewports data,and at last integrate all data in a same coordinate system. Thus,the model of object can be obtained.Registration for point clouds is so important in 3-D object modeling that it is the directly influencing factor to the final synthesis result and model precision.After summarizing the domestic and foreign scholar research results,this paper focuses on the registration of 3-D scanning data with contributions as follows:1.Analysis of space registration for 3-D laser scanning dataConsidering magnanimous data,large-scope and great number of scans in scanning of geographical scene,the method of space registration for 3-D laser scanning data is put forward. And it is verified by examples that three-step of sequence matching,overall matching and data integration is feasible,which can meet the requirements of space registration in geographical scene.2.Sequence matchingTo achieve the coordinate system's unification between two point clouds,the key lies in calculating seven transformation parameters,which require not less than three homonymous points.The accurate selecting of homonymous points has two situations:when labels exist in the scene,the label's central coordinate is obtained as the homonymous points;and when there are no labels in the scene,all points are firstly displayed in perspective projection's way in the OpenGL environment,and then the homonymous points are selected by man-machine interaction.Finally, based on the homonymous points,this paper has taken the indirect adjustment with conditions method to resolve the model and the analysis on errors is carried out.The results show that the two selecting method both meet the requirements of the accuracy of data modeling.3.Overall matching and data integrationOverall matching and data integration are main procedures in the space registration.In the process of sequence matching the errors are accumulated.In order to eliminate them,the global error distributing method is proposed based on the wire cloth station foundation.In addition,in order to delete the redundant data and simplify model,the paper proposes the data integration method based on the deleting redundant data on the spatial division in the overlap region.4.Data verificationBased on the researches above,the registration module of 3-D laser scanning data processing prototype system has been developed,which realizes the space registration for 3-D laser scanning data.The space registration experiments have verified that the method put forward in this paper could meet the reconstruction needs.