Research On Crucible Bubble Feature Detection And Measurement Method

The quartz crucible is mainly used for drawing monocrystalline silicon,which plays an extremely important role in the integrated circuit and the solar photovoltaic industry.When there are too many tiny bubbles in the transparent layer of the quartz crucible,it is easy to break in a long-term high-temperature vacuum environment,and the impurities in the bubbles will destroy the silicon crystal structure and affect the purity of the silicon ingot.This paper takes the bubbles in the transparent layer of the quartz crucible as the research object.It aims to perform quantitative analysis on the characteristic parameters of crucible bubbles and provide technical support for crucible quality inspection.Aiming at the problem of low contrast between crucible bubble outline and background,an improved dual-domain decomposition algorithm was revised.The filter template size is redefined and a soft threshold function is used to denoise high-frequency details.Experimental results show that the revised enhancement algorithm has a higher degree of detail assurance,a smaller light order error,and the enhancement result is more in line with human visual characteristics.Aiming at the problems of misdetection and missed detection in the extraction of crucible bubble contour,a crucible bubble contour extraction algorithm fused with Hough transform is proposed.Based on the "class" theory,the segmentation threshold is adaptively acquired,combined with morphological algorithms to fill the holes in the bubbles,and Hough transform is performed on the determined edge points to complete the contour extraction process.The experimental results show that the algorithm in this paper is more accurate for the pixel location of bubble contour.According to the four eigenvalue constraints,the preliminary classification of single bubbles and overlapping bubbles in the crucible bubble image is completed.Aiming at the problems of insufficient segmentation and over-segmentation in crucible bubble segmentation,a method of chain code marking corner points is proposed to locate the concave of overlapping bubbles.According to different overlapping types,different segmentation criteria are set to complete the segmentation process,and the overlapping bubble contours are fitted and corrected by constraints,and the distribution results of the crucible bubble diameter and number are obtained.The experimental results show that the detection accuracy of the algorithm in this paper is as high as 95.57%for the number of bubbles,which achieves a good segmentation effect.In order to verify the timeliness of the algorithm in this paper,a crucible bubble intelligent detection system based on ARM Cortex-A was developed.The algorithm in this paper is transplanted to the core control board to complete the image processing function.After testing,the designed system accurately locates the crucible bubble contour,and the detection time for a single image is only 0.1s.The results show that the algorithm in this paper can accurately and efficiently complete the measurement functions of quartz crucible bubble diameter,quantity and other parameters.It has certain theoretical significance and practical value in the automatic detection of quartz crucible.