Research On Three-dimensional Measurement Method Of Crucible Bubble Based On Video Image

Quartz crucible is an important container for the preparation of monocrystalline silicon.In the process of long-term contact with high-temperature molten silicon,the inner wall of the transparent layer dissolves,and the gaseous impurities in the inner wall enter the molten silicon,resulting in the degradation of the quality of monocrystalline silicon.In this thesis,quartz crucible bubbles are taken as the research object,and an image acquisition method of transparent layer bubbles of crucible is designed to complete the platform construction and video image acquisition.By detecting the bubbles in the image,the spatial position and radius of each bubble in the transparent layer of the crucible were obtained,and the number and gas holdup of bubbles in different depths were calculated to complete the three-dimensional measurement of bubbles in the transparent layer of the crucible.The image enhancement algorithm based on guided filtering is improved to solve the problem of low contrast between the bubble profile and background.The average variance weight is used to optimize the regularization parameters so that the filtered output image can keep the edge and smooth the flat region.The local adaptive gain parameters are used to complete the adaptive superposition of the detail layer in the edge region and the flat region.Experimental results show that the algorithm has a good overall enhancement effect on focusing bubbles.In the process of measuring two-dimensional plane focusing bubbles,it is difficult to identify and measure the focusing bubbles accurately because of the interference of non-focusing bubbles.In order to solve this problem,an improved image segmentation algorithm based on guided filtering is proposed in this thesis.Secondly,the radius ratio factor based on the maximum inner circle and the minimum covering circle is introduced as the circle detection and radius measurement algorithm,which overcomes the error detection problem caused by the complex edge of bubble and accurately measures the bubble radius.Then,a set of morphological combination was introduced to segment and reconstruct the focused adhesion bubble group.Finally,a bubble classification and screening algorithm based on image centroid neighborhood is proposed to further improve the accuracy of focusing bubble recognition and reduce the time complexity.The experimental results show that the average accuracy of the combined algorithm in this thesis is 95.53%,and the average error of radius measurement is 0.9%,which is 2.5% lower than the existing algorithm,and the detection speed is significantly improved compared with the experts.It is a difficult problem to accurately map multiple two-dimensional plane focusing bubbles into three-dimensional space bubbles during the measurement of three-dimensional space bubbles.In order to solve this problem,BIRCH algorithm based on threshold clustering feature tree was proposed to cluster XY coordinates of the center of focusing bubbles,and threshold promotion and segmentation factor strategy were used to solve the inaccurate clustering problem of BIRCH algorithm.The average clustering accuracy of this algorithm is98.6%,which is 2.7% higher than the original algorithm,and the time complexity is the same as the original algorithm.By choosing different crucible samples testing algorithm in this thesis,the following conclusions: In this thesis,the three-dimensional measurement of the transparent layer of crucible space bubble average recognition accuracy is 98.6%,the radius measuring average error is 0.9%,which can accurately describe the actual circumstances of the crucible transparent layer gas impurities,detection has higher reference value to the quality of crucible.