基于GPU的RFI抑制关键技术研究

射电天文观测过程中射频干扰（RFI）会通过直接或间接耦合方式进入接收系统，导致天体目标信号质量下降或信息丢失，对RFI进行准确识别和标记是减少数据损失和有效提高数据质量的关键。当前快速准确识别并抑制RFI已成为射电观测过程中急需解决的问题，项目针对新疆天文台南山25米射电望远镜（NSRT）在观测中遇到的干扰问题及大口径射电望远镜RFI抑制系统设备研制需要展开研究，拟采用FPGA+GPU实现天文信号的高速实时处理，建立以高性能GPU为计算核心的RFI抑制系统。研究并实现环形缓冲区算法，实现FPGA与GPU之间的高速数据传输；利用高性能GPU尝试并分析多相滤波、自适应滤波及基于人工智能算法的RFI抑制方法。拟将项目研究成果应用于NSRT，期望可以有效提高天文信号信噪比，并为正在建设的新疆110米望远镜RFI处理系统研制奠定理论和关键技术基础。

Radio frequency interference (RFI) will enter the receiving system through direct or indirect coupling during the radio astronomy observation, as a result, the signal quality of celestial target is reduced or the information is lost. Accurate identification and marking of RFI is the key to reducing data loss and improving data quality. At present, high-speed accurate identification and mitigation of RFI has become an urgent problem in the process of radio observation. The project is carried out research on the interference problems encountered in the observation of 25-meter Nanshan Radio Telescope (NSRT) and the development of RFI mitigation system for large aperture radio telescope, plan to use a next generation GPUs and FPGAs to realize high-speed real-time astronomical signal processing, and build RFI mitigation model based on high-performance GPUs. Investigate and implement the ring buffer algorithm to realize the high-speed data transmission between FPGA and GPU. High-performance GPU is used to implement polyphase channelizer, adaptive filtering and RFI mitigation methods based artificial intelligence algorithm. We plan to apply the research results of the project to NSRT, hoping to effectively improve the signal-to-noise ratio of astronomical signals, and it is lay the theoretical and important technical foundation for the development of Xinjiang 110-meter telescope RFI mitigation system.