Research On 3D Reconstruction And Mosaic Algorithm Based On Multi-focus Image Sequence

High-precision measurement of microscopic surface topography of objects has always been an important research direction in the field of measurement.It has broad application prospects and scientific research value in aerospace,military ship,precision engineering,micro-manufacturing,biotechnology and many other fields.In this paper,a method for measuring 3D surface topography of object based on multi-focus image sequence reconstruction is presented,aiming at achieving non-contact and high-precision measurement of object.The essence of this method is to recover the three-dimensional shape from the two-dimensional image sequence of the object and realize the three-dimensional digital measurement of the object.The three-dimensional reconstruction algorithm and planar mosaic algorithm of microscopic image sequences are mainly studied.Because of the problems of reconstruction algorithm and mosaic algorithm in processing microscopic images,a three-dimensional reconstruction algorithm based on non-downsampled wavelet transform and a homography matrix estimation method based on modified RANSAC algorithm are proposed.In this paper,three-dimensional reconstruction and mosaic algorithm of multi-focus image sequence will be used as a breakthrough to analyze the model surface distortion and mosaic error of multi-focus image sequence.This paper is divided into six chapters,mainly discussing the current status of Three-dimensional measurement method for multi-focus image sequence and the problems of three-dimensional reconstruction and image stitching algorithms.The improved algorithm is proposed for the existing problems,and the correctness of the algorithm is verified by the hardware and software measurement system.Specific work are as follows:(1)In the 3D reconstruction algorithm of microscopic image sequence,the 3D reconstruction algorithm based on non-downsampling wavelet transform is proposed.Based on the model quality evaluation,compared with the traditional 3D reconstruction algorithm,it can effectively solve the model surface distortion problem and improve the reconstruction quality of the model.(2)Estimation of homography matrix by traditional RANSAC algorithm to achieve mosaic of microscopic images.On the basis,aiming at the problem that the traditional RANSAC algorithm has many iterations and the accuracy of the homography matrix is low,a homography matrix estimation method based on modified RANSAC algorithm is proposed to improve the accuracy and efficiency of the homography matrix estimation.(3)Build height calibration and XY plane calibration experiments,Based on the rigorous and complete model quality evaluation,the 3D model reconstruction experiment,the multi-view 3D model plane splicing experiment,the model surface parameter extraction experiment and the roughness extraction experiment were constructed.(4)A complete hardware system with relevant algorithms is established.It can realize the acquisition of the microscopic image sequence in the step distance of 2?,and transmit the collected image data to the PC through the image acquisition card to realize the processing of the digital image.