使用二分类神经网络对GONG多普勒图像进行异常检测
Anomaly Detection for GONG Doppler Imagery Using a Binary Classification Neural Network
AI总结：
本文针对NISP背面地震图的质量保证问题，提出一种二分类卷积神经网络，用于自动筛选GONG站点图像是否应纳入背面处理流程，从而高效提升最终地图质量。
摘要：
美国国家太阳天文台综合天气计划（NISP）的产品之一是背面地震图，它展示了太阳未被观测面的磁活动。这些基础地图的制作始于2006年，此后已被证明是追踪无法从地球表面直接观测到的太阳活动的宝贵工具。对太阳活动区的持续追踪使空间天气预报员能够监测可能对地球产生更大经济和社会影响的关键太阳事件。为了改进这些地图，由NSF资助的“宇宙之窗”项目（WoU）正在开展若干步骤。其中一个步骤是改进对GONG网络各站点采集图像的质量保证措施，这些图像被用于生成背面地图。为此，我们设计了一个二分类神经网络，以确定哪些站点图像应被纳入生成最终产品地图的背面处理流程，哪些不应被纳入。该卷积神经网络是一种高效且计算上经济的方法，可显著提升NISP计划当前所产出的背面地图的质量。
英文摘要：
One of the products of the National Solar Observatory's Integrated Synoptic Program (NISP) is the farside seismic map which shows the magnetic activity on the unobserved side of the Sun. The production of these rudimentary maps began in 2006, and they have since proven to be a valuable tool in tracking solar activity which cannot be directly observed from the earth's surface. The continuous tracking of solar active regions allows space weather forecasters to monitor critical solar events which may have larger economic and societal impacts here on Earth. In an effort to improve these maps, several steps are underway through the Windows on the Universe project (WoU) funded by the NSF. One of these steps is to improve the quality assurance measures for the images collected at individual sites throughout the GONG network and is used to develop the farside maps. To this end, we have designed a binary classification neural network to determine which of these site images should and should not be included in the farside pipeline that produces the end product maps. This convolutional neural network is a highly effective and computationally efficient method of significantly improving the quality of the farside maps currently produced by the NISP program.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
