Research On Compressed Ultrafast 3D Imaging Based On Interference Fringe

Three-dimensional(3D)imaging is the technology fusing sensors and computing methods together to obtain the surface profile of target objects.Among them,3D imaging based on interference stripes has the advantages of high precision,high resolution,noncontact,non-injury and other advantages.The 3D imaging technology is widely used in manufacturing,aerospace,medicine,cultural heritage protection and other fields,providing more accurate and reliable measurement methods and analysis tools for various fields.However,the traditional 3D imaging cannot meet the requirement of ultrafast temporal resolution 3D detection.It is possible to overcome this problem by combining the Compressed Ultrafast Photography(CUP)and 3D imaging together to achieve ultra-high temporal resolution 3D imaging.Based on this consideration,this thesis develops an interference striped 3D imaging technology with CUP to establish a compressed ultrafast 3D imaging framework.The specific research work is as follows:(1)Establish a 3D imaging system based on Mach-Zehnder interferometer,and perform camera calibration,interference fringe image analysis and 3D imaging calculation on the imaging system.Collecting deformation interference stripes for three-dimensional imaging can obtain its 3D cloud model.(2)Combining the 3D imaging system in(1)and the CUP system,a compressed ultrafast 3D imaging method is proposed.The imaging mathematical model,reconstruction method of the compressed ultrafast imaging system are studied.Existing reconstruction algorithms for the CUP system mainly include Two-step Iterative Shrinkage/Thresholding(TWIST)Algorithm and Augmented Lagrangian(AL)Algorithm.As a promising image reconstruction technology,the Plug-and-Play Fast and Flexible Denoising Network(PNPFFDNET)provides an efficient solution with high image denoising quality.We employ this technology in the compressed ultrafast 3D imaging for the first time in this thesis.Experimental results show that the PNP-FFDET algorithm can reconstruct interference fringe images of various dynamic processes in the CUP system.Compared with the traditional TWIST and AL algorithms,the PNP-FFDET algorithm can improve the peak signal-to-noise ratio of reconstructed fringe images by at least 10 d B and the structural similarity by at least13%,indicating higher image reconstruction quality.(3)In the 3D measurement process,various noises caused by environmental factors will degrade the accuracy of the resulting point cloud.To improve the efficiency of point cloud denoising,a method based on image segmentation is proposed.Comparative experiments show that this method effectively removes point cloud noise without the complex calculations in 3D space.Experimental results show that the point cloud denoising accuracy of the proposed method can reach 99.974%.