Pattern recognition for MEMS images of surface topography using wavelets

Micro-Electro-Mechanical Systems (MEMS) have been growing in interest in recent years. Due to the small size of the MEMS, the traditional method of metrology measurement seriously affects the parameter of the object being measured, and high accuracy metrology cannot be acquired. MEMS microstructure images of surface are most composed of a number of steps, grooves and slots. Pattern analysis and recognition of these microstructures with linear feature is one of the key problems in metrology and testing technology of Micro-Electro-Mechanical-System (MEMS). Effective detections of these components play an important role in simplifying feature model in pattern recognition and pattern match. Traditional Linear feature detectors based on pixel processing each by each may fail to detect out lines in image with low SNR. A fast discrete beamlet transform and a novel method of linear feature detection are proposed, which can detect lines with any orientation, location and length. Experiment results prove the efficiency of the method proposed even in image with very low SNR.