Research On 3D Reconstruction Technology Based On Binocular Telecentric Structured Light

Three-dimensional reconstruction technology is an important means to obtain the geometric shape of objects.The three-dimensional reconstruction technology based on structured light has received extensive attention and research due to its non-contact,high-precision and high-speed characteristics.This paper studies the 3D reconstruction technology of structured light for small-sized objects.The paper replaces the pinhole lens in the traditional binocular structured light system with the telecentric lens,and build a high-precision binocular telecentric structured light 3D reconstruction system.Key technologies such as telecentric lens calibration method,efficient codec design,and stereo matching in structured light systems are discussed.The main research contents of this paper are as follows:（1）The traditional telecentric camera calibration method needs to take multiple different poses plane target photos for calibration.During calibration,the target needs to be manually moved,which is time-consuming.In addition,the high-precision flat target is expensive and has a fixed size,which is not able to meet the calibration needs of different magnification lenses.Aiming at the shortcomings of the traditional telecentric camera calibration method,this paper proposes a telecentric camera calibration method based on a virtual plane target.The method generates a series of virtual targets through predefined transformations.Then,use the telecentric camera to capture the orthogonal projection of the virtual target on the screen,calculate the homography matrix from the virtual target plane to the camera imaging plane,and decompose the homography matrix to obtain the parameters of the telecentric camera.Experimental results show that the method achieves accurate calibration results and eliminates the need to manufacture and manually move the physical planar target multiple times.（2）In order to solve the problem that the traditional Gray code-assisted phase-shift coding technology projects more pictures of the Gray code and the decoding error at the boundary of the Gray code causes the phase unwrapping error,a gray phase-shift code encoding method based on third-order dislocation is proposed.First,the second-order Gray code is extended to the third-order Gray code,so that the number of projected Gray code pictures is reduced from［log₂^f］to［log₃^f］,where[·]is the round-up function,f is the number of sinusoidal fringe periods.A phase-shift fringe-assisted ternarization third-order Gray code picture method is proposed during decoding,which avoids the traditional gray code thresholding segmentation method that needs to project additional black full White pictures are needed.On this basis,a third-order staggered Gray code coding strategy is proposed.The image of the third-order Gray code is decoded to obtain the initial phase order,and then an additional third-order Gray code pattern is projected to shift the Zhang generation and the initial phase order to the left by 1/3 period and right-shifted 1/3 period misalignment order,the three orders are integrated to avoid phase unwrapping errors at the edge of the codeword.The experimental results show that compared with the traditional Gray phase-shift coding method,the projected image is reduced by 30%,while the robustness of the traditional Gray code-assisted phase-shift method is maintained,and the phase unwrapping error at the boundary of the Gray code of the traditional method is overcome.（3）Aiming at the problems of many projected images and long matching time in the traditional bidirectional phase stereo matching method,a phase stereo matching method based on epipolar correction under telecentric optical path is proposed.This method can perform epipolar correction on the images captured by the left and right telecentric cameras,so that the points to be matched are on the same line as the imaging points on the left and right cameras,so it is only necessary to project the vertical Gray code phase shift coding pattern on the surface of the object,and then use the phase information and template matching to determine the sub-pixel precision matching point.The experimental results show that compared with the traditional matching algorithm,the number of projected coded pictures is reduced by half,the matching time is faster,and the accuracy is higher.（4）Developed structured light 3D reconstruction software on the Qt platform,integrated the algorithms of each link of reconstruction,and used OpenMp technology to accelerate the computation involved in the algorithm in parallel.The software can complete tasks such as telecentric camera calibration,camera parameter setting,structured light encoding and decoding,point cloud generation,point cloud visualization,etc.