Research On 3D Reconstruction Algorithm For Persevering Sharp Feature

In recent years,the research and application of 3D reconstruction technology has been paid attention to the fields of computer vision,computer graphics,reverse engineering,archaeology,medical image processing,digital entertainment,etc.Based on the reverse engineering,this paper studies the 3D reconstruction method with sharp edges,and hopes to reconstruct a model that can maintain the surface features of the original object.In order to make the follow-up measurement operation more accurately and splicing operation more conveniently after acquiring the 3D model,it is necessary to obtain more accurate 3D point cloud data in the initial stage of 3D reconstruction.We use the method of raster scan to acquire the 3D coordinate of point cloud directly,which compared to the traditional way,whether in time or accuracy are greatly improved.Later,we analyze the 3D point cloud data and aim to be able to reconstruct object more quickly and accurately.In this paper,we research the key technology of 3D reconstruction algorithm and the specific contributions of this paper is as follows:In the stage of preprocessing 3D point cloud,we extend image smooth method of 2D image to 3D space to solve the noise produced in point cloud.And we use different smooth method to solve different noise;for the outlier may arise in the process of collecting point cloud,we come up with an improved clustering method based on radius search to find and remove the outlier points.In the stage of merging point cloud,to solve the problem of dislocation and redundancy of point cloud during collection process,we use iterative closest point algorithm for point cloud registration and put forward a method based on neighborhood search and boundary detection to remove the point cloud in duplicate area.We proved that our method can effectively improve the accuracy of point cloud fusion.In the stage of reconstructing point cloud,in order to solve the problem the edge or corner area had in the reconstruction process can't maintain the structural characteristics of origin object,this paper presents a method of surface reconstruction based on feature detection to preserve the sharp feature.This method finds out the approximate location of feature edge by feature detection on origin point cloud to improve search speed.And then proposes a selfadaptive iterative method to update point coordinate.And in this way we can build the featurepreserving 3D model more quickly and accurately.This paper uses point cloud with rich sharp edge to carry out an experiment on the above method,the experimental results show that compared with other surface reconstruction methods which maintain features,our method can get 3D model on the sharper edge in fewer iterations case.