Research On Model-based Building3D Construction Method

Remote sensing technology in the contemporary fast development, themilitary and civilian reflect its importance. Through the different sensor platform,can gain a lot of remote sensing data, the data in the accurate extraction ofground targets3D information, target for three-dimensional structure plays animportant role. As people more and more high to the requirement of3D building,platform, long phase, data from sensors have their respective advantages anddisadvantages, through the complementary information, can make up for theinadequacy of different conditions, further complete the build results,information is the best. And build on that target than the construction of thewhole scene will have higher requirements, target top surface and edgeinformation is particularly important.This paper established the three-dimensional information decoding module,provide important data for subsequent three-dimensional building support. Underdifferent models were compared and analyzed, on the one hand, the features andadvantages and disadvantages, and study the effect of different conditions on theparameters. Main research rigorous sensor model and general sensor model andits improvement, outside the interior orientation parameters and solving themathematical model of directional parameters, and then complete the3Dinformation of settlement, on the other hand, comparing the differentconvergence Angle effect on calculating precision, to determine their position ofsensor to obtain the best data, reduce the error, reduce the error transfer in thesystem of data applications of its important role in the future.Later, mainly research based on the method of establishing typical topologymodel of three-dimensional. Considering the three dimensional informationdecoding process, due to get the data or match error, may lead to the number ofeffective calculating point, LTD. In this case, the establishment of a typical targetmodel library, the complex target simplification, according to the differentparameters of different model input, finally realizes the solid build. Targetmainly concentrated in building, oil tank and the simple bridge. When obtainingan effective calculating point is large, can top surface model is set up, usingsurface fitting to complete3D information added. At the same time verifyunderstanding calculate point number and distribution has important effect oncalculating precision.Finally, a multiple-neighborhood filtering (MNF) method is proposed for building3D-information recovering. Optimization of digital surface models(DSM) is the most necessary part in3D building reconstruction. Points could notbe determined in all areas owing to matching failure. MNF mainly focuses onbuilding rather than whole scene by other methods. Building has fewer pointsthan whole scene and its edge is harder to recover. As MNF is not aninterpolation, it smoothly extrapolates to recover edge of building. We can builda gridded surface with the values at a set of nodes. Experiments show MNF couldgreatly improve completeness of3D-information. Completeness afteroptimization is stabilization around85%. MNF has a good robustness.