Abstract:

The surface attributes of the semiconductor are studied by using the artistry of gray scale processing, image segmentation, edge detection, and mathematical morphology. Deficiency cannot be completely avoided in the production process of the semiconductor. These bumpy and irregular areas may directly affect the growth of the crystal, consequently diminishing the percent of pass and practicality of the device. The semiconductor SEM (Scanning Electron Microscopy) images were processed by divergent kinds of edge detection operators (Sobel, Canny and Log).  The edge detection results of divergent kind operators are respectively serial and parallel, and the clearest and complete results obtained. The results of image processing technology can be used to quantitatively analyze the flaws of the semiconductor surface, to achieve the semiconductor surface uneven area of automatic positioning and identification. It can effectively prevent the formation of artificial detection errors, contribute to measurement accuracy and increased efficiency.