Font Size: a A A

The Research Of Indoor Scene Texture Mapping Method Based On Model Segmentation

Posted on:2019-03-05Degree:MasterType:Thesis
Country:ChinaCandidate:Y SuFull Text:PDF
GTID:2428330566498134Subject:Control Science and Engineering
Abstract/Summary:
The rapid development of computer vision promotes the application of realistic 3D modeling in our life.Texture mapping is the most important means to reinforcement realistic on the vision model.Indoor scene reconstruct effect depends mainly on the accuracy of the texture and the fusion effect.However,the traditional texture mapping method use the same algorithm to the whole model which will produce local texture dislocation.Based on above this paper puts forward a indoor scene texture mapping solution based on scene segmentation.We improve the whole texture effect through the study of indoor scene 3D model scene segmentation,and texture information fusion algorithm based on the depth of field label.Solve the indoor scene texture-model mismatch,eliminate the texture aberration,completed panoramic deep shade.The main research contents and methods include:First,we set up the indoor scene model reconstruction system.Design model data collection method.Using hardware to get depth information and texture information for indoor scene 3D model reconstruction.According to the characteristics of indoor scene,set different projection classification to fit different part of the model.Secondly,analysis the effect of reconstruction,using probability graph to segment model and texture into foreground and background,then mark the result separately.Change the original physical corresponding relation between model and texture,correct the mismatch of texture.This process innovative implement through reverse projection depth model to texture photo,transform three-dimensional segmentation to twodimensional problem.In this way,we achieve data dimension reduction,make texture and model combined with complementary way,improved the boundary corresponding accuracy between texture and model.Finally,use SIFT to achieve the optical optimization of global texture which already be transform into HSV color space.Segmentation boundary grid,initialize each texture weight of every single grids,then recombination the weight of neighborhood texture to ensure the coherence of fusion area.Integrate the final weight which will be project to the surface of the model in order to promote the final effect of 3D texture mapping.
Keywords/Search Tags:Indoor scene, 3D reconstruction, Scene segmentation, Reverse mapping, Texture fusion
Related items