Abstract: Based on the HL-2A Tokamak optical imaging system, an algorithm for plasma boundary reconstruction is designed, which has the advantages of fast operation speed and outstanding robustness to various noise signals. Firstly, the optical image is processed by the basic algorithm of digital image processing, then the plasma boundary coordinates in the image are extracted and fitted. Finally, the actual coordinates of plasma boundary are obtained by coordinate transformation, that is, the reconstructed plasma boundary. Comparing the optical reconstructed plasma boundary with the last closed flux surface by EFIT, the average errors of the high and low field side boundaries are about 8.5mm at midplane, and the overall average error is 11mm. The average time to identify each image on the experimental hardware equipment and software platform is 3ms. The proposed algorithm is expected to be used in plasma configuration identification and control under the current experimental results. It is consistent with EFIT qualitatively and quantitatively. In addition, it is not affected by magnetic field, and has excellent stability and robustness.

Key words: HL-2A tokamak, Peak brightness line, Boundary reconstruction, Image recognition

摘要： 基于HL-2A托卡马克等离子体可见光成像系统设计了用于等离子体边界重建的算法，具有运算速度快、对各种噪声信号具有较好的鲁棒性等优点。算法首先运用数字图像处理的基本算法处理可见图像，然后提取图像中的等离子体边界坐标并进行拟合，最后经过坐标变换得到等离子体边界的实际空间坐标，即重建的等离子体边界。将重建的等离子体边界与EFIT重建的最外闭合磁面进行对比，中平面强、弱场侧位置的误差小于8.5mm，边界总体平均误差小于11mm，在实验中的硬件设备及软件平台上识别每幅图像的平均耗时为3ms。该算法在目前的实验结果下有望用于等离子体位形识别及控制中，与EFIT在定性和定量上具有一致性，并且不受磁场影响，具有良好的稳定性和鲁棒性。

关键词: HL-2A, 亮度峰值线, 边界重建, 图像识别