Research Of Modeling And Measuring Based On Images

In the field of urban reconstruction, it is necessary to reconstruct3D model ofinteresting target from its2D images. The approach of Image-Based Modeling(IBM) is to restore geometric model of object and scene by using lots of known2Dimages. Compared with traditional modeling methods of using modeling softwareor3D scanner, the advantages of IBM are low cost, high fidelity, fast reconstruction.Although the research of IBM is wide, but how to improve modeling accuracy andaccelerate modeling speed is still the research focus of many researchers. Theapproach of IBM involves in some sub-issues, such as stereo matching, surfacereconstruction. The accuracy and efficiency of these sub-issues are to be improved.Deep research of these sub-problems can contribute to further development of IBM.Image-Based Measuring is the embodiment of IBM applied in practice,Image-Based Virtual Space represents the space by using images, and so how tocarry out the measurement in Image-Based Virtual Space is worth of particularconcern.Focus on the research of modeling and measuring based on images, thework of this dissertation is listed as follows:(1) The proposition of an improved adaptive support-weight matching algorithmbased on multi-factor. Traditional matching algorithm is studied. Multiplefactors affecting matching results are analyzed and added to matchingalgorithm in a reasonable manner in order to achieve better matching results.(2) The proposition of a3D surface reconstruction algorithm based on non-uniformdiscrete set of points. By using the role of directional points in constructing theminimum spanning tree, a new approach of finding neighbor points is proposed.In order to achieve high-quality surface reconstruction of non-uniform discreteset of points, constraints are introduced to impose restriction on the subsequenttriangular grid step.(3) The implementation of a measuring and locating approach based onImage-Based Virtual Space. Against inefficiency and difficulty for measuringand locating operation of objects in the field of urban traffic management, byusing Image-Based Virtual Space, the length measuring and locating of trafficsigns can be more accurately.