Research On Key Technology Of 3D Measurement System Based On Structured Light Projection

With the flourishing development of technology and information industry,3D measurement technology has gradually entered into people's production and life,the acquisition and measurement of 3D data of objects has been gradually applied to such fields as reverse design,industrial manufacturing and testing,medical rehabilitation,education and teaching,etc.3D measurement technology of active projection structured light is favored for its non-contact,high precision and full field.However,the nonlinear error introduced by digital projection equipment and phase error optimization problems,as well as the calibration of cameras and projectors,3D reconstruction,and phase unwrapping for absolute phase recovery are key technical problems that the system needs to overcome.Although some scholars have developed the ideal 3D measuring system,the conditions for using the 3D measuring system are harsh,and the objects that can be measured are also very limited,the results of the complex,discontinuous,and more shadows conditions are not satisfactory,the accuracy of the algorithm also needs to be improved.This paper studies the above-mentioned key technologies of 3D measurement system based on structured light projection,and summarizes the advantages and disadvantages of related algorithms in 3D measurement systems through experimental verification,and combines some problems faced in practical applications,proposes some improvements and new ideas,and built a complete 3D measurement system of active projection structure light.The following are the main work contents and related technical improvements of this article:1.In this paper,the phase-shift algorithm and phase-unwrapping algorithm for absolute phase calculation are studied,through experimental verification and comparison,the unwrapping capabilities and disadvantages in different situations of three commonly used space-based unwrapping algorithms,Goldstein branch cut,DCT,and least squares,are analyzed.For the insufficiency of the spatial unwrapping algorithm,this paper experimentally verifies the advantages and disadvantages of the time phase-shifting method,and combines the nonlinear problems of the system,leading into new thinking on the unwrapping algorithm.2.In this paper,the absolute phase error optimization algorithm is studied.The phase error caused by the Gamma nonlinearity of the system is eliminated by a nonlinear correction experiment.The sinusoidal of the fringe pattern before and after correction,unwrapping phaseand error variation are analyzed experimentally.An improved unwrapping algorithm for tri-band fringe projection is proposed,the algorithm combines the traditional tri-band fringe projection method and the eight-step phase-shift method,only three fringe patterns are projected at each low-frequency group,the unwrapping phase result can be used to determine the fringe level,no need to be exact,the high-frequency fringe map group needs to project 8 fringe patterns,and the phase shift difference between every two fringes is pi/4,the wrapped phase of the high frequency fringes is calculated by the wrapped phase formula of the eight step phase-shifting method,then use the fringe level determined by the group of low-frequency fringe patterns to unwrap,and the unwrapped phase is the final absolute phase.The improved algorithm avoids the lack of spatial unwrapping and reduces the amount of projected images.Experimental results show that the improved algorithm can reduce the nonlinear error of the system to a certain extent and improve the accuracy of the unwrapping.3.In this paper,camera calibration and projector calibration algorithms are studied,the calibration experiments are carried out respectively,and the experience of calibration is summarized.The algorithm of mark point extraction is studied,and proposed an improved SUSAN checkerboard corner detection algorithm.The improved algorithm obtained the difference of three pixels gray sum between two groups of nonadjacent characteristic directions,combining the threshold and difference threshold to extract the corner set.Moreover,using the symmetric distribution of the corner set to determine the centroid coordinates of the corner set as the coordinates of the corner of the position,the coordinate values can be directly sub-pixel accuracy.Experiments show that the proposed algorithm can extract corners effectively and quickly,and the reprojection error is within 0.3 pixels.The extraction results can be used for camera calibration.4.This article builds a complete 3D measurement system and realizes 3D reconstruction of the front cover interior of BMW cars,the pre-sintering ball of iron-carbon micro-electrolytic fillers,artwork models,gasoline saw muffler accessories,etc.Stripe generation and synchronization triggering device and camera triggering software are designed.