Research And Implementation Of Diamondquality Inspection Technology Based On Deeplearning

Synthetic diamond is produced under the condition of high temperature and high pressure which simulatethe natural environment.In the production of synthetic diamond,the main factors affecting the quality of synthetic diamond are temperature and pressure.Under a certain pressure,the growth rate of the synthetic diamond varies in different,directions with the change of the synthesis temperature field,as a result,the macroscopic appearance of synthetic diamonde particles is different.At present,the temperature field in the process of synthetic diamond production cannot be accurately controlled.Generally,the quality of synthetic diamond can indirectly reflecte whether the synthetic temperature of diamond is suitable.In th diamond production,the crystal shape of diamond particles is usually observed artificially under a microscope in a graphite rods,and compared with the standard crystal shape template to determine the synthesis temperature high or low,and then the synthesis temperature process curve was adjusted.But the artificial observation method is inefficient and the consistency of detection quality is not good engough,which will cause errors in production process adjustment.This paper prsents the research and system realization of automatic detection technology of diamond particle quality by machine vision.Combining with the synthetic technology of a diamond company in Zhengzhou,the system adopts an industrial camera(CCD)to collect the sectional image of synthetic graphite rod section,and detect the quality of diamond particles based on the detection technology of deep learning technology.Firstly,the growth mechanism and synthetic equipment of synthetic diamond are introduced,then,based on the theory and technology of deep learning,Faster R-CNN,improved YOLOv3,improved SSD and other deep network frameworks were used to conduct intelligent classification of diamond particle images.By comparing the results of FPS,mAP,ACC,the improved YOLOv3 network framework with relatively simple structure and faster detection speed was finally selected to realize AI detection of the grain shape,particle number and particle area of synthetic diamond.In this paper,the detection system developed by the research has achieved an accuracy rate of 96.44% in the detection of diamond particle shape in graphite rod,which is consistent with the actual situation in the detection of diamond particle number and diamond particle area.number and area of synthetic diamond particles.Among them,the detection system The machine vision detection system usingAI technology has the advantages of high precision,fast speed,and real-time feedback,which overcomes the limitations of manual detection.