Research On Key Technology Of Micro-Motion Measurement Based On Computer Microvision

Modern science and technology is rapidly developing to tiny, ultra-precision field from micron, submicron into the nanoscale stage. With the fast development of the technologies such as micro/nano manufacturing, microelectronics and biomedicine, micro-motion measurement is strongly demanded. Novel high-precision measurement methods are hence in urgent need for micro/nano manufacturing and biomedicine. Computer micro-vision method has been increasing attention in the field of precision measurement for its flexible, rapid, non-contact, precision and high degree of automation. This thesis focuses on micro-motion measurement with high accuracy using computer micro-vision approach. The main contributions of this thesis are listed as follows:First, the computer micro-vision measurement system scheme and its hardware components are in-depth studied, the effects of light source on the imaging quality are systematically analyzed, the properties of the acquired images and the noises introduced by the computer microvision system are explored, and the main approaches to attenuating these noises are analyzed.Second, Aiming at the problem of illumination variation, noise and large displacement that may badly influence motion estimation precision in micro-motion measurement based on computer microvision. Three kinds of high-precision robust motion estimation algorithm are proposed:(1) According to the imaging model of micro-vision, a robust multi-scale micro-motion measurement algorithm based on homomorphic filtering is proposed. First, a method of homomorphic filtering for image enhancement is used to correct the uneven brightness of micro-vision image and enhance contrast. Then biweight function is used to automatically adjust the weight of data with different residual error and to remove those data with excessive residual errors, and a multi-scale pyramid is employed to accurately estimate the motion vector by an iteration gradually from coarseness to fine. Experimental simulation show the new algorithm has good robustness, it can effectively weaken the influence of uneven illumination in micro-motion measurement based on computer micro-vision and reduce the interference of outliers caused by noises, and the accuracy of micro-motion measurement is improved.(2) According to the model based on image structure, an integrated approach of robust motion estimation based on constancy assumption of monogenic phase was proposed, which use for illumination changes and low SNR. The monogenic signal is the first rotation invariant two-dimensional analytic signal. Its phase information includes structural information and has the advantage of being robust to illumination changes and noise interferences. Therefore, the method combines multi-scale pyramid iterative methods and robust estimation methods for motion estimation would further improve the motion estimation accuracy under illumination changes and low SNR conditions. Experimental simulations verify the good performance of the algorithm.(3)A motion estimation algorithm based on monogenic curvature tensor and digital image correlation is designed by using correlation method. In the algorithm, first 1D and 2D structural information of the image is extracted by monogenic curvature tensor to improve motion estimation accuracy under uneven illumination by using the advantage of image structure being invariant to illumination change. Then, motion displacement of integer pixel is obtained by correlation method. On this basis, the gradient-based sub-pixel displacement estimation algorithm combined with robust methods to reduce the effect of noise on the accuracy of motion estimation and obtain sub-pixel displacement vector. Experimental simulations verify that the algorithm can not only be suitable for rigid body motion estimation in uneven illumination and noise conditions, but also be suitable for displacement field estimation of deformation objects.Finally, using the precision positioning stage as measurement objects, translation and rotation micro-motions are measured using the proposed methods. Experiment results verify the theory and methods in this thesis are feasible and effectively achieve the high-precision motion estimation in illumination variation and noise conditions.