Geometric Feature Extraction And 3D Reconstruction Of Artificial Objects

Image feature extraction is a basic problem in Computer Vision.Unlike natural scenes,the artificial objects which is design by geometric graphics has more imformation about geometric features and less image texture.Therefore,the accuracy of artificial objects' s geometric feature extraction and description relates to higher level visual tasks such as recognition and 3D reconstruction.In this paper,we have done some researches on the geometric feature extraction and 3D reconstruction of artificial objects.To represent an edge map in the form of an edge block,this paper analyzes the connected domain property of edge points,proposes a labeling model and search strategy to aggregate edge points into edge blocks.It is verified that the edge block can retain the edge graph information completely.This paper proposed a novel line extraction algorithm base on edge block,which aims to slove the problem that the line segment is eliminated due to the lack of gradient information in the current line detection algorithm.On the basis of the edge blocks,the adjacency matrix is used to calculate the position information between the edge blocks,and the depth-first search algorithm is presented to obtain the initial edge chain which called path in this paper.With the geometric characteristics of the line,this paper filters the path set to complete the line feature detection process.The experimental results show that the algorithm can effectively extract the linear features in various scenes,and our algorithm can keep the geometry information of the object completely.Using the proposed algorithm to extract the line features of artificial objects,and these features will be input into the Line3 D framework to obtain the model reconstruction of artificial objects.The experimental results show that the 3D result based on line features has more complete structural information.