Design Of Single Crystal Silicon Rod Measuring System Based On Image Processing

Single-crystal silicon is a crucial material in fields such as IT and solar energy.To produce high-quality single-crystal silicon using the direct pulling method,it is necessary to measure the crystal diameter accurately and in a timely manner to ensure production efficiency.Using vision technology for detection in single-crystal silicon manufacturing is a good non-contact technique.However,in terms of accurate image measurement,multiple challenges still exist.In this study,we propose a novel single-crystal silicon diameter measurement system that can meet these industrial requirements:A ROI extraction image pre-processing algorithm based on adaptive thresholding is proposed.The impact of noise on the image is analyzed and the image is denoised using median filtering.The adaptive thresholding,using a combination of intelligent optimization algorithms and an improved OTSU algorithm,separates silicon rods and background in different single-crystal silicon scenarios by calculating the appropriate threshold value,utilizing image connectivity properties for separating the silicon rods from the measured image and circling out the ROI.By eliminating background interference,the subsequent image algorithm processing can become simpler and more effective.In this study,we propose an improved image edge detection algorithm that combines pixel-level and sub-pixel-level edge detection.The pixel-level edge algorithm is employed to improve the efficiency of edge detection,whereas the improved Zernike edge detection algorithm is used to enhance accuracy.Various conventional edge detection algorithms like Sobel,Laplacian,and Canny are used for the rough extraction of single-crystal silicon rod edges.Through comparative tests,the mathematical morphology algorithm is found to have some advantages in terms of extracting silicon rod edge details and reducing image processing time.For further processing of the edges extracted through pixel-level edge detection,an improved sub-pixel edge detection algorithm based on Zernike moments is proposed.However,the relative grayscale threshold of the Zernike algorithm needs constant adjustment.The algorithm determines the relative grayscale threshold by using the maximum between-cluster variance of the image segmentation threshold.Different extraction algorithms are used to handle measurement issues in different stages of diameter algorithm measurement,aiming to enhance the accuracy and efficiency of diameter algorithm measurement.For incomplete circular images in the isodiametric stage,a lower edge extraction algorithm is used to measure the sub-pixel coordinates of the image edge.For the shaking amplitude is relatively large but there is a small bright circle lead crystal stage and no obvious aperture release shoulder stage,the three feature points of the leftmost,rightmost and bottommost are extracted,and the circle center coordinates and radius are determined by the algorithm of three points to determine the circle.4)Finally,the diameter measurement system was developed and designed in C#to measure the diameter of single-crystal silicon rods.The experimental results show that the absolute error of the diameter measurement of silicon rods in the lead and shoulder release stages is 0.1mm,while the maximum In summary,the image processing-based silicon rod diameter measurement system proposed in this thesis has some value for production practice and future silicon rod diameter measurement.