中文摘要：

提出一种将计算机微视觉与鲁棒多尺度运动估计算法相结合的测量微运动的方法。设计一种用于微运动测量的鲁棒多尺度运动估计算法。在该算法中，根据滤波器的不同特性，在多尺度金字塔的不同尺度层采用不同的滤波器估计运动。这种分层估计的方法不但可以提高测量精度，而且可以加快测量速度。分析计算机微视觉测量系统的组成，并对设计的微视觉测量系统进行标定。以精密定位平台为测量对象，采用计算机微视觉与鲁棒多尺度运动估计算法相结合的方法测量了精密定位平台的平面微运动。该方法用于测量2000nm左右的微运动时，最大测量偏差达到了12．5nm。试验结果表明，这种将计算机微视觉与鲁棒多尺度运动估计算法相结合的方法能够实现高精度的平面微运动测量。

英文摘要：

An integrated approach for micro-motion measurement that incorporates robust multiscale motion estimation algorithm into the computer microvision is presented. A robust multiscale motion estimation algorithm for micro-motion measurement is proposed. In this algorithm, according to the different properties of filters, motion at each level of the pyramid is estimated using different gradient filters. This hierarchical estimation technique can not only improve the measurement accuracy, but also accelerate the measurement speed. The composition of computer microvision measurement system is analyzed, and the designed microvision measurement system is calibrated to measure micro-motion. Taking the precision positioning stage as measurement objects, the in-plane micro-motions of the precision positioning stage are measured by using the proposed combined approach. The maximal bias of this new approach reaches 12.5 nm for motions near 2 000 nm. Experimental results show that the new combined method can measure the in-plane micro-motions with high accuracy.