3D Reconstruction Technology Using Monocular Multi-focus Images

Microstructural components have been widely applied in equipment manufacturing,biomedicine,aerospace and other fields.It is of great significance to realize the 3D surface measurement of microstructure components to guarantee the reliable application.Therefore,it has great application value to study the high-precision 3D reconstruction method for small objects.Through the analysis of the development of 3D reconstruction methods,the depth from focus method was further studied in this paper.The focusing method changes the distance between the measured object and the focusing system.The depth information of the surface topography could be obtained by implementing the image processing algorithm on the acquired serial images to identify the sharp areas in each image.In this case,the 3D reconstruction of the surface topography could be accomplished.Due to uneven illumination,hardware equipment problems and et al,the image quality could be poor which would produce the error of image sharpness judgment.A PCA-based improved Retinex image enhancement method was proposed to solve the problem.Compared with other image enhancement algorithms,the results show that the proposed method effectively compensate for the uneven illumination,increase the detail information,and retain the gray distribution of the original image.In order to obtain the sharp position of each area of the object surface contained in the image,a reliable image sharpness evaluation function is necessary.Through comparative analysis of results from the different image sharpness evaluation functions,the Tenengrad function was chosen in this paper.Furthermore,an image block partitioning strategy based on statistical analysis was proposed,which could automatically determine the reasonable block size and benefit the accurate and fast extraction of the sharp areas to obtain the 3D measurement in multiple depth of field.A 3D reconstruction experimental device was built up according to the measurement principle.The device could obtain the 3D reconstruction of of a region of the object surface by collecting one group of sequential images.A standard step was tested.The measured height of the step is 0.206 mm,which is 0.013 mm different from the standard height of the step.And the standard deviation of the experimental data was 0.003 mm.In addition,the slope plane surface and the top surface of the sphere were measured.The 3D morphology of the slope and the sphere were gathered.The effectiveness of the proposed method was verified.