Abstract: A novel method for automatic three-dimonsional(3D) stitching of microscopic images of micro electro mechanical systems (MEMS) microstructure is proposed. After the microscopic images of MEMS microstructure are captured by white-light vertical scanning interference microscopy(WVSIM), the microscopic image of MEMS microstructure which contains complete information of MEMS microstructure with large field and high resolution is obtained by the proposed method. It could meet the demand of the analysis and evaluation of functional characteristics of MEMS microstructure. By analyzing the principle of WVSIM, the process of 3D stitching microscopic images of MEMS microstructure is decomposed to x-y stitching and height correction. Scale invariant feature transform(SIFT), which is frequently used for 2D image registration, is applied to 3D stitching of microscopic images of MEMS microstructure, so that it is smart and robust. The strict alignment is unnecessary and the problem of the automatic stitching of microscopic images of MEMS microstructure with arbitrary shape with high precision is solved by the method. The experiment result shows that the lateral accuracy could be reached 0.8 μm and the vertical accuracy is smaller than 1nm.

Key words: Automatic three-dimonsional(3D) stitching, Micro electro mechanical systems(MEMS) microstructure, Scale invariant feature transform(SIFT), White-light vertical scanning interference microscopy(WVSIM)