Research On Large Depth Of Field Microscopy And 3D Measurement Based On Focus Variation Technology

With the development of precision machining and micro-devices,micro-surface topography measurement technology has also made great progress,and its application is increasingly broad and deep in many fields,such as materials science,aerospace and so on.Although the existing non-contact measurement methods have their own advantages,they can not measure large dip slopes and different material surface.The shape from focus(SFF)technology that has emerged in recent years is expected to solve this problem.This paper focuses on the large depth of field microscopic and three-dimensional zoom measurement technology,the main work is as follows:Based on the principle of optical small depth of field and SFF measurement,an experimental system was built,including a small depth of field optical imaging system,motion control system and image acquisition system.sequence images with different degrees of focus were taken.Due to the unstable experimental environment,such as the vibration of the stage and the change of illumination,the images may mismatch.Therefore,it is necessary to register the sequence images so that they are aligned in the geometric position to facilitate subsequent processing.The registration method based on feature and gray information is mainly analyzed.The structural similarity(SSIM)is used to quantitatively evaluate the registration quality.Registered image and reference image was compared by overlapping display to make comprehensive measurement of the registration effect.the SURF feature-based registration method is determined considering the requirements of speed and accuracy for sequence image registration.Aiming at the problem of depth of field limitation in microscopic imaging,multifocus image fusion studies were performed to obtain large depth of field microscopic images.The fusion experiments are carried out from three aspects: spatial domain,pyramid transform domain and wavelet domain.The advantages and disadvantages of different methods are analyzed from the objective evaluation value and the effect of the fusion image.It is concluded that fusion effect based on transform method is better than the spatial domain.Although some details of the fusion image based on wavelet domain are lost,the overall effect is better than pyramid fusion method.In order to select the appropriate focus measure function,the influence of different focus evaluation operators and window size on the measurement results is compared and analyzed.the preliminary depth map is obtained according to the three-point Gaussian interpolation near the focus peak.At the same time,after median filtering on the depth map,reliability analysis is performed to remove some error points and improve the overall surface structure.The large depth of field fusion image is combined with the depth map to reconstruct the three-dimensional surface topography of the measured object,and the error analysis of the standard gauge block is used to verify the feasibility of the algorithm.