Practical Application Of Image Processing In Single Crystal Silicon Diameter Detection

Single-crystal silicon is a very lively nonmetallic element and it is mainly used in semiconductor materials and solar photovoltaic industry. Because silicon substrate is formed from the single-crystal silicon of equal-diameter stage, so the control of the diameter is an important part. The measure of diameter of single crystal silicon mainly adopts the non-contact detection. Single-crystal silicon of equal-diameter stage is taken by the CMOS camera when it is rotating.What we get is a part of video and image of single-crystal silicon because of being blocked by silicon rod.In the part of image processing, a part of the information of aperture should be obtained in order to calculate the diameter of single crystal silicon.The image taken by CMOS camera with some minor noise will affect the clarity and quality of image, so it need smoothing noise. In this paper,what is used is a kind of easy to implement adaptive directional median filters algorithm. It is able to filter the noise area instead of filtering the whole image and it could select automatically suitable window,so it can restrain noise as well as protect the detail of the image edge. There are so many background elements in the image and the aperture is the highest brightness.So the way of analysising the histogram is used to get the best threshold and the image is dealed by binarization. The result shows that the halo around silicon rod and aperture have been isolated and the result of image segmentation is quite good.The actual precision requirement of the edge is0.1mm, but the edge made by classical edge operator can only reach pixel and it is not up to the actual requirement after conversion. Thus,the algorithm of sub-pixel edge detection is introduced to further improve the precision of the edge. The precision of the edge is about0.013mm by using the sub-pixel edge detection algorithm based on Gauss fitting and it has reached the expected requirement. The outer edge of the single-crystal silicon is only retained because of the requirement of calculation.Because of the high melting point of the single-crystal silicon,the camera can only shoot by overlooking outside of the observation window and what we get is an incomplete elliptical. A kind of improved method is put forward by this paper that is about diametral calculation. First, make the incomplete elliptical into a complete circle, then look for the location of maximum diameter on this curve. Finally,the pixel value of the diameter is got by using the principle of sub-pixel. The dip angle θ is directly introduced by this algorithm and its advantage is that it can get the diameter by calculation without the value of θ.