Object Plane Structure Modeling Based On Local Distribution Feature Image Matching

3D reconstruction based on image matching is a common 3d modeling method.Under the condition of sparse sampling with large viewing angle,the integrity of the 3D model reconstructed by traditional methods is not high due to the lack of image information.In this case,the geometric features of the image change greatly,the number of feature point pairs obtained by image matching is small,and the 3D model information obtained by point based 3D reconstruction is seriously missing.Therefore,this method matches the image according to the local distribution characteristics of the image,uses the plane fitting method to segment the plane structure of the object,and uses the plane as the unit for 3D modeling,so that a more complete object model can be constructed using a small amount of image information.The main work of this thesis is as follows:Sparse sampling means that the perspective of the image changes greatly.In order to make full use of the local distribution characteristics of the same kind of feature points and improve the image matching effect under the conditions of large viewing angles,this thesis presents an image matching method based on the local distribution characteristics.This method simulates the process of human eye vision from far to near objects.First,the image is divided into several color regions according to the pixel values,and then the color features are extracted as the overall features of the object.Then,the outline features of the image are extracted as the detail features from the global image to ensure the integrity of the feature points under the conditions of large viewing angles.In feature matching,first select the reference point,use the distribution distance to build the image relative coordinate system for fast matching,and then local matching in the way of region growth,correct and complete the feature point pairs,improve the matching rate of feature points,and ensure the angle invariance of features.The comparison results show that,compared with classical image matching methods,the feature matching rate of image matching methods based on local distribution characteristics is higher under large viewing angle conditions.Although this method extracts fewer feature points,it matches more correct feature pairs.In order to construct as complete a three-dimensional model as possible with a small amount of image data,this method uses planar information to construct object models,and a lightweight threedimensional modeling method is obtained.To ensure the integrity and accuracy of plane information,this method uses local distribution characteristics to improve the plane fitting method based on image matching.In order to improve the discontinuity of plane distribution and the integrity of the model,this method calculates the edge lines of plane structure based on the local distribution characteristics of the outline feature points,which improves the plane integrity.In order to extract more complete and accurate edge lines and improve the accuracy of plane fitting,a three-step outline extraction method is presented in this thesis.The comparison results show that,compared with traditional threedimensional modeling methods,this method can use a small number of images to construct a more complete planar structure model of objects,and it simplifies the modeling steps,reduces the computational load,reduces the requirements for processors,and expands the scope of use of three-dimensional modeling technology.