Extraction Of Semiconductor Surface Image Parameters Based On Image Processing

Dislocation density is one of the important index to evaluate the quality of the crystal, crystal surface topography photo shows the distribution of crystal dislocation, counting for the dislocation numbers, then the dislocation density could be get. Dislocation of traditional statistical method is selected the center of a dislocation of gray level as a benchmark in the image, and set a threshold value, if the gray level is in the area that we settled the unicom are counted as center of dislocation. But this method depending on the selected benchmark, the threshold value is different, the size of the statistical result differs very big, plus some of the useless area is also in the gray level area we settled, it also falls within the scope of the statistics, the result and the actual deviation is bigger. In order to get accurate results, often need to artificial count, and the original image due to the interference of background color, corrosion area distribution chaos, easy to cause visual fatigue, reduce the efficiency of artificial statistics. In view of this situation, this article from the several aspects to carry out the research work:At first, the color images which be collected in the lab will be taken into gray photo as a pretreatment. Then the picture, which is the single wafer of semiconductor In P material dislocation etch pits of topography, would be optimized by the Sobel operator, Canny operator and Daubechies wavelet Symlets, Haar wavelet processing methods to get the edge detection.Secondly the basic theory of watershed algorithm is introduced, and is pointed out the disadvantages and the insufficiency. Then adopting the method of marking on the improvement, and using mathematical morphology algorithm to optimize the processing result, separate the center of the dislocation segmented from the complex background, and get the binary image of the central region of dislocation. And using the method of euler number to count for the dislocation in the image after segmentation centers for automatic statistics.Finally, the Matlab GUI is designed to make the algorithm becomes visual program, through the procedures for the calculation of several other image, the effect of the algorithm is verified, the statistical efficiency improved significantly. And the results of the second chapter and the third chapter is combined, it has carried out the further result to get the edge of the original image detection.