Image Recognition Method Of Critical Seeding For Sapphire Crystal Growth Via The Kyropoulos Method

This article is based on the project of science and technology bureau of Jinzhong City,Shanxi Province and the enterprise development project(Grant No.201803D03111003,Grant NO.193220027-J).The laboratory cooperated with Zhongjujingke Co.,Ltd.to solve the problem of low seeding efficiency during the sapphire growth process.Based on the basic theory of seeding,a new method for identifying critical seeding status was proposed,a visual detection model of critical seeding was established,and the feasibility of the model was analyzed and verified.The research results show that identification of critical seeding status is the key to improve seeding efficiency.The seeding is the first step in sapphire growth.Due to the strict periodicity and continuity of the atomic arrangement of sapphire single crystals,the seeding technology directly affects the quality of sapphire single crystals.Defects generated during the seeding stage will spread to the entire crystal.However,the existing artificial seeding method mainly relies on the experience of the seeding technicians,and the judgment of the seeding status is subjective.At the same time,the seeding temperature is about 2050℃,which is difficult to observe with naked eyes in high temperature and strong bright environment.The above subjective and objective factors led to low efficiency of artificial seeding.At present,the seeding theories and corresponding technical indicators have not yet been formed.This paper combines the image processing technology,regards the regular flow of the melt free surface as the basis for seeding judgment,and forms a new method for identifying critical seeding status.It also proposes that critical seeding status identification is the core of seeding technology optimization.Aiming at solving the objective problems in artificial seeding,this article sets up a visual detection platform which consists of an industrial CCD,industrial lens and several neutral attenuation filters.The platform is used to collect free surface images.Aiming at solving the subjective problems,this paper establishes a visual model of critical seeding based on the spoke formation principle on the melt surface,the seeding image recognition method and the critical seeding feature identification method.The visual model of critical seeding includes image processing model,velocity model,position model and seeding state recognition model.According to the gray distribution characteristics,noise types and spoke structure of the seeding,Open CV library is called by C++ compiler language to realize the filtering,skeletonization and feature extraction.The seeding speed model and seeding space position model are fitted to analyze the spoke movement characteristics.The standardized results are input into the critical seeding judgment model to realize the critical seeding status identification.The visual model of critical seeding was validated and compared with the Observation Convergence Seeding(OCS)method in the accuracy of seeding detection.The results prove that the method in this paper can effectively improve the anti-interference ability of the seeding status identification under the actual thermal field conditions.Therefore,accurately identifying the critical seeding status can improve the seeding efficiency,which has guiding significance and application value for the research and development of the automatic seeding technology.