The Method And Realization On Annotation Optimal Labeling Of The 3D Real Estate Scene

3D Real Estate Scene Annotation Labeling Problem(3DRESALP)takes an important role of the visualization of semantic information in 3DRES.According to the differences in memory rendering mechanism,display platforms and the patterns of annotation labeling,3DRES is divided into Large-scale Real Estate Scene(L-3DRES)and Small-scale Real Estate Scene(S-3DRES).L-3DRES formed by many 3D features mainly be used for displaying application scenarios.It generally uses Real time cache rendering mechanism.The annotation in L-3DRES,generally adopt Screen Space Annotation,requires labeling efficiently rather than express detail.Consequently,L-3DRESALP focuses on building and optimizing the Annotation Display Policy.The main contents are:(1)Analyzing Spatial cognition law of eye,building the levels of annotation model of regional real estate scene from the level of expression way.(2)Building the Annotation Level Display Policy and expressing all the names step by step,including cadastral subarea names,district names,road names,mark building names and other buildings' names.S-3DRES has less features,which uses directly loading into memory mechanism.It is mainly applies to Specific property object displaying.The Annotations of S-3DRES are commonly used as Object Space Annotation and focus on detail view require high-quality.The research and implementation of S-3DRESALP based on the rules,algorithms and quality evaluation function of annotation.The main contents of this part are:(1)Researching the annotation configuration rules.Combined with 2D map annotation configuration theory,summed up 3D annotation configuration requirements,so to get S-3DRES features(such as:3D parcels)annotation configuration rules.(2)Designing and Building annotation labeling algorithms for S-3DRES,which contains perspective transformation matrix and inverse perspective transformation matrix,GRID algorithm,Genetic algorithm,the evaluation function of annotation quality based on GRID.The experiment demonstrates that the Strategies and a family of algorithms of L-3DRESALP?S-3DRESALP proposed in this paper are effective.The Annotation Level Display Policy designed in this paper improves evidently with the deficiency of too dense to unable to identify.The Annotation Labeling Algorithm focused on S-3DRES is capable for multi-view 3D parcel annotation labeling and reaches better effects than the direct annotation labeling effect of Supermap,ArcGIS platforms.The loss of annotations changes 55%and the overlap of annotations lessens 100%,what's more,the conflicts of annotations cuts down 48%.Setting different annotation placement rules according to different 3D Geographic Scene,this algorithm can be used in 3D annotation labeling widely.