Automatic Inspection of Typical Microstructure Defects Based on Image Processing Techniques

Abstract

Micro-structural defects influence the performance of MEMS devices. For effective quality monitoring and assurance in mass-production of MEMS devices, it is necessary to conduct automatic defect inspection. In this paper, an approach based on image processing for automatic microstructure defect inspection is presented. In the approach, an Influence Region Template is built based on a reference microstructure image while it preserves the framework and restrains the tiny distortion of the reference, then the test microstructure image is compared with the Influencing Region Template for quality evaluation. By the approach, typical defects such as blocks in trenches or channels, transfixions between trenches or channels and missing of some part and so on, are detected and classified automatically, while some distortions which do not affect the function of the microstructures are ignored. Testing results show that the approach can conduct a high quality assurance while it can endure a given distortion tolerance of microstructures.