Study On 3D Scanning Measurement Technology Of Structured Light

It is difficult to reconstruct the 3D shape of complex components in the process of industrial measurement field.This paper introduces a handy 3D scanner equipped with binocular CCD cameras and cross semiconductor laser which is designed independently.On the basis of binocular stereo vision technique and structured light measurement technique,we have deeply studied and analyzed about the principle of full-field self-positioning scanning system and difficulties of 3D reconstruction.A new non-coded circle self-positioning algorithm for reconstructing the 3D profile of complex object is proposed,and built the hardware and software to verify the accuracy and reliability of this method.Through data processing and error analyzing,it is proved that this system has highly measuring precision and speed.The main research contents and achievements are as follows:1.This paper proposes a new technique of non-coded circle self-position innovatively.We can identify and distinguish with the circles accurately in the epipolar direction by the uniqueness diameter of circles.Based on the image circles matching exactly,we can realize the registration of corresponding points by the topological invariant relation between the circles and solve the difficulty how to mosaic the discontinuous point cloud data under the different coordinate transformation.2.Thinning and distinguishing cross laser line.Due to the highlight feature of the laser stripe,it can be extracted by the adaptive threshold segmentation algorithm.The existed light stripe center extraction methods have been analyzed,choose the center of gravity method to thin the light stripe.The ordinary cross laser line segmentation algorithm will not be able to distinguish accurately in the case of intersection point fails.Based on the binocular stereo calibration and rectification,we can calibrate the cross structured-light separately,then distinguishing it with the correct pair of image points calculating 3D points which is in the structured-light plane.3.The whole process of image processing is divided into four parallel stages by using the pattern of pipeline.The four stages are as follows: image acquisition,cross laser stripe extraction,non-coded circle recognition and matching,point cloud reconstruction.Theoretically,the processing time can be reduced to 1/4 if we can adjust to balance four stages of image processing for load balancing between threads.