Research On Face Matching Algorithm In 3D Reconstruction Of Buildings

With the improvement of camera quality,the development of computer vision science and technology is becoming more and more mature,more and more tends to digitize,and radiates to more fields,such as scene model reconstruction,intelligent city,antiquities model digital storage and display,Visualization of map scene,virtual electronic assistant medicine,3D online game industry and hybrid reality all require high-resolution and high-precision three-dimensional measurement model.Three-dimensional reconstruction is to calculate the spatial data of three-dimensional model by single,multiple two-dimensional image information or some feature point set information obtained by computer.Three-dimensional reconstruction of buildings has been widely used in urban planning,disaster monitoring and virtual tourism.It has been an important research topic in computer vision,pattern recognition and mapping for a long time.One of the mainstream methods of building 3D reconstruction is to use the reconstruction method based on image sequence and monocular camera to restore the 3D model of the general scene by continuously fusing the feature point set extracted from the image pair,and using the single-eye camera to restore the 3D model of the general scene.This kind of method requires more scene images in order to extract a large number of matching point sets.Therefore,the time complexity of this method is very high,and the wrong matching points will seriously affect the accuracy of 3D reconstruction.Considering that the building models are usually composed of multiple planes,a multi-plane reconstruction method based on multiple homography is proposed in this paper.Firstly,the building image sequence is extracted and matched by SIFT feature points,and the wrong matching points are eliminated by RANSAC algorithm.Then the set of matching points is transformed to get the set of characteristic points of different planes,from which the high-setting region of the plane is fitted,and then the principal plane is determined from the separated plane,and the auxiliary plane is corrected.Classification finds different planes,uses high-confidence sets to represent each plane,and finally is flattened by the main structure The intersection lines and detection lines are used to constrain the high confidence set,expand the plane region and reconstruct the plane part of the scene.Then the three-dimensional model of the object is obtained by repeated plane reconstruction.In this paper,from the perspective of surface matching reconstruction,the multi-planar model is reconstructed by fewer image sequences of frames.The experimental results show that the reconstruction method based on plane matching can quickly and accurately reconstruct the model composed of multiple planes,especially for weak texture images.