Research On Correspondence Search Of Digital Speckle 3D Reconstruction

With the development of the society,3d reconstruction technology is increasingly used in industrial production and various fields of social life.Among of many 3d reconstruction technologies,digital speckle 3d reconstruction get more and more attention because of the advantages that only one pair of speckle images can reconstruct the 3d data of the measured object.However,this method needs to calculate the corresponding function of a large number of matching points when searching for the corresponding points.This process computes a huge amount,which obviously affects the efficiency of 3d reconstruction.In this paper,we focus on improving the search efficiency of the corresponding points of digital speckle reconstruction.The main work includes two parts:Part one,to acquire three-dimensional reconstruction of the measured object by only one snap based on digital speckle and binocular stereoscopic vision.Firstly,extract the speckle area of the original speckle image,reducing the search range and the matching error rate of the corresponding point.Secondly,using the projective rectification to transform the tilted conjugate epipolar line into the same horizontal line,it provides great convenience for searching the correspondence and improves the search efficiency.Then,using the digital speckle correlation method to get the correspondence and remove the inaccurate match point according to parallax continuity.Finally,optimize the sub-pixel and reconstructed the cloud data of the object.Part two,a novel correspondence searching method based on gray constraint proposed,which at the basis of projective rectification and disparity constraint.This method removes most of the candidate corresponding points according the difference of candidate correspondence points and the measurement point,hence reduce the correlation computational frequency and shorten the correspondence searching time.The experimental results show that,when the threshold is set to 20,70% of the candidate correspondence within parallax eliminated successfully,the searching time by about 30% of only using parallax constraint,efficiency about 2 times.When the electrical signal synchronized to acquire the speckle image,threshold can further reduced to 12,the searching time by about 21% of only using parallax constraint,efficiency about 4 times.Compare to only using parallax constraint,the proposed method can greatly improve the efficiency of integer pixel correlation searching,and the efficiency increase along with the window size of correlation.