| With the development of the mirco/nano science and technology,the micro/nano materials have been widely applied in the high-tech fileds such as chip manufacturing,electronic packaging,biological medicine,etc.Due to the size effects of the micro/nano materials and structures,they can be easily deformed,cracked and then fallen in failure under the multi-field coupling loads such as force-electrical-magnetic-thermal.Therefore,it is vital to make precise micro/nano-scale 3D surface measurement for understanding mechnical performance,analyzing the failure mechanisms of micro/nano materials and guidng the design and maching of mirco/nano system.In recent years,with the advances in precision measurement technology under micro/nano scale,there are various micro/nanoscale 3D measurement methods.Among them,methods based on Scanning Electron Microscope(SEM),also called 3D SEM,held a series of advantages including: high efficiency,non-contact,wide measuring range and insensitivity to surface roughness.Therefore,3D SEM has attracted many scholars' attention.However,a SEM is designed for visualization,not for metrological studies,and the image formation is biased by several image disturbances.There are still some problems in the area such as SEM imaging model and calibration,image distortion correction,3D reconstruction algorithms.To solve above mentioned key issues,this dissertation is dedicated to establish a general parametric imaging model for SEM and propose a flexible calibration method;to investigate the pattern of SEM image distortion and correct them;to develope an efficient and local 3D measurement method based on the mapping between disparity and depth and propose a adaptive SfM framework for dense and global 3D measurement.Based on the above research content,a complete set of 3D SEM surface measurement theory and technology system can be formed.The research work is concluded as follows.Aiming at the problems of the considerable differences about SEM imaging model and the chosen of critical magnification,a smooth general model is built for characterization of SEM imaging characteristics without relying on any hypothetical conditions.The proposed smooth general model is based on the continuity constraint of the SEM imaging process,and the radial basis function is introduced to model the continuity constraint which essentially represents mapping between the image point and the corresponding line in space.By the use of the proposed smooth model,the relationship between magnification and imaging model is clarified,and the nature of imaging process is revealed.The SEM imaging model is demonstrated in a generalization and visualization way at the first time.The visualization results validate the former assumptions about the relationship between the imaging model and the magnification at the first time.The accuracy experiment results demonstrate that smooth general imaging model for SEM is practically effective and its principle is much closer to the SEM imaging theory than traditional ones'.The proposed method provides a new way of exploring the mathematical model of SEM,and it holds important theoretical and technical value.The image formation is biased by several image disturbances.The causes of SEM image distortion are complicated and there is no significant regularity.As a result,it is really difficult to apply traditional parametric distortion models to correct them.To solve the above issues,a SEM imaging distortion model is presented considering the magnification change.Starting from the source of the distortions,time-correlated and position-correlated models are presented to express drift and spatial distortions,respectively.According to imaging characteristics under different magnifications,two types of spatial distortion calibration methods are developed individually based on grid target and speckle target.Based on the above theories and innovative methods,the SEM imaging distortions can be well corrected under variational magnifications.The experiment results demonstrate that the proposed distortion modeling and correction method can correct the image distortions effectively.The unexpected displacements decrease from ±4 pixels to ±0.5 pixels and the virtual strains are reduced from ±6000? to ±1000?.The unexpected influence caused by distortions is eliminated.The measurement accuracy can be improved.In order to improve the efficiency of measurement process,it is significative to rapidly reconstruct local 3D topography with a small amount of images.To address the problem,an 3D measurement method based on the mapping between disparity and depth is proposed,also called D2D-SEM.The mapping between disparity and depth is conducted to achieve the rapid resolving and reconstruction.The epipolar rectification is introduced to ensure the accuracy and stability of the disparity map and final results.Digital Image Correlation(DIC)is adopted to calculate the dense disparity map.Integrating the above techniques,the proposed D2D-SEM can recover the local topography of high quality with only two SEM images.Comparing with the Digital 3D Microscope and the Laser Scanning Confocal Microscope,the accuracy and effectiveness are verified.The experiment results demonstrate that the proposed D2D-SEM can make a SEM capable of accurate 3D measurement based on only two images.It hold unique advantages including low cost,insensitive to surface roughness and high reflective surface.In the view of how to use multiple SEM images to perform complete 3D measurement,an adaptive 3D reconstruction framework is proposed based on SfM pipeline(ASfM-SEM).Under the framework of multi-view reconstruction theory,the core algorithm of 3D reconstruction is revised to adapt the parallel projection model.By applying the proposed smooth general model,a flexible switch can be realized based on variational magnifications of SEM to enaure high-precision three-dimensional reconstruction.To solve the issues about image noise and feature matching,the feature-based and area-based method are combined to accurate and robust feature matching.Base on above techniques,ASfM-SEM can achieve complete 3D measurement.The comparison experiment show that the ASfM-SEM holds the same accuracy and reliability with the digital 3D microscope.Severl samples are measured by the ASfM-SEM,it can be concluded that it can realize accurate and complete 3D measurement of the measured samples,especially for the measurement requirements for fine details and occlusion situation.In conclusion,the dissertation addresses the major needs of micro/nano-scale 3D reconstruction,measurement and characterization.It conducted an in-depth study in 3D SEM including: the nature of SEM imaging model has been revealed,the pattern of SEM image distortion has been investigated,an efficient and local 3D measurement method is developed and an adaptive SfM framework for complete 3D measurement is proposed.The successful implement of this dissertation has important scientific significance and potential applications that it can provide theoretical fundamentals and techniques for micromechanical testing of chip manufacturing,electronic packaging etc. |
PDF Full Text Request