3 D Building Reconstruction Based On Stereo Vision Technology Research

In recent years,3D reconstruction technique based on the stereovision is one of the most important fields in computer vision. The reasonwhy the technique can be a hot topic is that3D reconstructiontechnique could realize the3D reconstrucion for2D image quciklywithout expensive equipment and cost. The demand for3D productand real sense of the world is perfectly fulfilled.3D reconstructiontechnique is widely applied in our daily life: In medicine industry,through the perspective scan, the3D model of viscera can bereconstructed, which is extremely beneficial for patients’ conditionanalysis; In military field, by photographing the actual combat site, thebattlefield model can be rebuilt to realize the virtual war scene; Inconstruction industry, we can simulate the architecture models into thevivid city scene through the real3D technology to give the audience abetter perspective.This paper mainly focus on the3D reconstruction of the architectureissues. The structures of this paper are listed as follows:Firstly, we improve the classic Harris corner detector methods andpropose a new Harris self-adaptive threshold corner detectionalgorithm based on enhanced wavelet detailed coefficients,whichmakes the corner positioning much more precise and even, whichcould be beneficial for the subsequent matching operation.Secondly, for more accurate matching purpose, we propose amethod that combines Harris-Laplace scales with SIFT algorithm to getinitial matching pairs and then uses RANSAC robust algorithm to get theaccurate matching pairs, which could be effective in removing themismatches and estimating the fundamental matrix.Thirdly, after the discussion of the existing camera calibrationmethod, we use Zhang calibration method, which is situated between the traditional calibration method and self calibration method, toachieve a more accurate calibration result through raising the precisionof corners from pixel to sub-pixel. Combining precision of the matchingpairs and the fundamental matrix is quite helpful for estimating theprojection matrix more percisely.Finally, in3D reconstruction, we use SFM algorithm to obtain the3Dspace discrete points.The experiments verify the effectiveness of the algorithm propsed inthis paper in the3D reconstruction field.