Research And Implementation Of Optical Boundary Reconstruction On EAST

Optimized plasma discharge requires accurate control of parameters such as plasma configuration,and the external measurement data can be used to reconstruct the plasma edge without directly measuring these parameters,thereby providing optimized control parameters.At the same time,physical diagnostic data analysis requires accurate and reliable plasma edge reconstruction results as a reference.Therefore,it is of great significance to conduct accurate and fast EAST plasma edge reconstruction studies.With the rapid development of computer vision technology and high-speed CCD camera software and hardware technology,the use of high-speed computer vision technology as a fusion device diagnosis has become an important trend.The EAST device is the world's first fully superconducting tokamak device.With the continuous improvement of the operation level of the device and the annual upgrade of the host and diagnostic system,it is possible to reconstruct the plasma edge through visible light diagnosis.In order to reconstruct the plasma edge by visible light diagnostic data,it is first necessary to extract the plasma edge in the visible light image.In this paper,two edge extraction image processing algorithms are proposed,which are edge extraction algorithm based on two-steps(offline algorithm)and local explicit edge extraction algorithm based on region of interest(real-time algorithm).The algorithm results are compared vertically and horizontally with the commonly used edge extraction algorithm and the edge extraction algorithm results of other devices.The results show that the real-time algorithm can achieve the fast requirements and precision requirements for real-time plasma shape control.The advantage of off-line algorithm is that it has higher precision and does not need to manually divide the region of interest in the image.In order to complete the optical reconstruction of the plasma configuration,a set of algorithms for image coordinate system of the plasma edge to the world coordinate system is developed.Firstly,the nonlinear distortion model in the calibration process is introduced.The Tsai algorithm is introduced in detail as the calibration method in the reconstruction algorithm.At the same time,considering the special experimental conditions of the large tokamak device,a reconstruction algorithm based on regression analysis was developed,which is a multi-linear regression analysis or BP neural network combined with Harris corner detection reconstruction algorithm.Accurate calibration of the optical path before each round of experiments can directly give the coordinate system conversion homography matrix at the beginning of the experiment.At the same time,after the device vibrates,the homography matrix combination can be corrected by regression analysis.The addition of the Harris corner detection algorithm can realize the image coordinate value acquisition of the feature points in the image.Based on the algorithm development,a computer hardware system platform for real-time reconstruction of EAST visible light image plasma edge is built.And based on the existing optical path conditions,the real-time experimental results of the spring experiment of 2019 spring are displayed.The results show that the system platform equipped with the reconstruction algorithm of this thesis can complete the plasma edge extraction of the mid-plane in 100 microseconds.After the offline EFIT result is used as the regression analysis to obtain the reconstructed homography matrix,the final reconstruction result and the offline EFIT contrast error The range is within 2cm.Finally,the algorithm and data transmission acceleration of the real-time optical reconstruction diagnosis system are introduced.