基于容器技术的天文应用软件自动部署方法

平方千米阵列即将开始建设,各子工作包也进入关键设计评估阶段。基于云与容器技术是平方千米阵列科学数据处理器未来可能采用的平台技术。针对超大规模海量数据处理面临的天文应用软件快速部署、运行与实测要求,充分考虑天文应用软件运行环境复杂、云计算环境下超大规模计算集群部署困难等问题,系统研究并给出了一种使用容器技术的天文应用软件通用自动部署方法。以目前较为常用的可见度函数校准软件SAGECaL为例,首先分析了SAGECaL的相关特性和分布式部署方面存在的困难,进而给出了基于容器技术的SAGECaL分布式集群的自动部署方法。实验结果表明,自动部署方法极大地提高了SAGECaL分布式集群的部署效率,满足项目组承担平方千米阵列科学数据处理器相关测试工作所需要的基础平台部署与切换等需求,同时也为其它天文软件在云端的快速部署与执行提供了有益的思路。     

Test area of the SAGE survey

Sky surveys represent one of the most important efforts to improve developments in astrophysics,especially when using new photometric bands. We are performing the Stellar Abundance and Galactic Evolution(SAGE) survey with a self-designed SAGE photometric system, which is composed of eight photometric bands. The project mainly aims to study the stellar atmospheric parameters of ～0.5 billion stars in ～12 000 deg2 of the northern sky, which mainly focuses on Galactic astronomy, as well as some aspects of extragalactic astronomy. This work introduces the detailed data reduction process of the test field NGC 6791, including the data reduction of single-exposure images and stacked multi-exposure images, and properties of the final catalog.     

A PCA approach to stellar abundances I. testing of the method validity

The derivation of element abundances of stars is a key step in detailed spectroscopic analysis. A spectroscopic method may suffer from errors associated with model simplifications. We have developed a new method of deriving the various element abundances of stars based on the calibration established from a group of standard stars. We perform principal component analysis(PCA) on a homogeneous library of stellar spectra, and then use machine learning to calibrate the relationship between principal components and element abundances. By testing with spectral libraries S4 N and MILES, we find that our procedure provides good consistency when spectra from a homogeneous set of observations are used, and it could be expanded to stars with quite a wide range of stellar parameters, with both dwarfs and giants. Moreover, we discuss the four key factors that have a significant impact on the results of derived element abundances,including the resolution of the spectra, wavelength range, the signal-to-noise ratio(S/N) of spectra and the number of principal components adopted.     

The galaxy luminosity function in the LAMOST Complete Spectroscopic Survey of Pointing Area at the Southern Galactic Cap

We present optical luminosity functions(LFs) of galaxies in the~(0.1) g,~(0.1) r,~(0.1) i bands, calculated using data in~40 deg~2 sky area of the LAMOST Complete Spectroscopic Survey of Pointing Area(LaCoSSPAr) in the Southern Galactic Cap. Redshifts for galaxies brighter than r = 18.1 were obtained mainly with LAMOST. In each band, LFs derived using both parametric and non-parametric maximum likelihood methods agree well with each other. In the0.1 r band, our fitting parameters of the Schechter function are φ_*=(1.65 ± 0.36) × 10~(-2) h~3 Mpc~(-3), M_*=-20.69 ± 0.06 mag and α =-1.12 ± 0.08,which agree with previous studies. Separate LFs are also derived for emission line galaxies and absorption line galaxies. The LFs of absorption line galaxies show a dip at~(0.1) r ~18.5 and can be fitted well by a double-Gaussian function, suggesting a bimodality in passive galaxies.     

Low-mass and high-mass supermassive black holes in radio-loud AGNs are spun-up in different evolution paths

How supermassive black holes(SMBHs) are spun-up is a key issue in modern astrophysics. As an extension to the study in Wang et al., here we address the issue by comparing the host galaxy properties of nearby(z 0.05) radio-selected Seyfert 2 galaxies. With the two-dimensional bulge+disk decompositions for the SDSS r-band images, we identify a dichotomy in various host galaxy properties for radio-loud SMBHs. By assuming that radio emission from the jet reflects a high SMBH spin, which stems from the well-known Blandford-Znajek mechanism of jet production, high-mass SMBHs(i.e., MBH 107.9 M⊙)have a preference for being spun-up in classical bulges, and low-mass SMBHs(i.e., MBH = 106-107 M⊙)in pseudo-bulges. This dichotomy suggests and confirms that high-mass and low-mass SMBHs are spun-up in different ways, i.e., a major "dry" merger and a secular evolution respectively.     

深度卷积神经网络支持下的遥感影像飞机检测

针对YOLOv3算法对小目标检测较差及出现较多漏检的问题,本文提出了一种优化的YOLOv3算法.首先使用K-means算法计算出与数据集相适用的锚框;其次将扩张卷积引入到YOLOv3网络,用来增强网络高层的感受野,改善小目标的检测效果;然后使用深度可分离卷积取代YOLOv3网络残差模块中的普通卷积,可减少计算量,从而得到一种新型卷积神经网络结构;最后在数据集上进行对比试验.结果表明,优化的YOLOv3算法能够检测出更多目标,降低漏检率,相比于YOLOv3算法,其召回率提高11.86%,F1-score提高2.99%.     

基于GPS新型L5信号的地表雪深反演研究

利用GPS多路径反射信号测量地表雪深具有全天候和高时空分辨率的特点,因此其可作为一种代替气象站监测雪深的新手段。然而,先前大多数研究仅使用了GPS L1和L2C波段信噪比数据探测积雪深度。为验证新型的L5信号在雪深反演方面的优越性,本文阐述了GPS-R技术反演雪深的原理,利用Lomb-Scargle周期图法所处理的受积雪表层影响的信噪比数据计算了频谱振幅强度,通过获取频谱特征值与天线高度的关系求解雪深值,最后分别与L1反演结果和实测雪深数据进行了对比。试验结果表明：与现有的GPS-R测量雪深结果相比,利用新型的L5反射信号反演地表雪深的精度更佳;采用GPS-R技术探测雪深对把握测站区域内的雪深变化情况和淡水资源储量具有重要价值。     

一种基于GA-BP-MC神经网络的高铁桥墩沉降预测模型

提出一种基于马尔科夫链修正的遗传BP神经网络预测模型（GA-BP-MC）,利用遗传算法的全局寻优能力初始化BP神经网络权值和阈值,初步建立GA-BP神经网络预测模型,结合马尔科夫链的无后效性修正模型预测值,形成高精度GA-BP-MC神经网络变形预测模型。结合高铁桥墩沉降数据,分别与BP神经网络、GA-BP神经网络预测模型进行对比,结果表明,该预测模型精度最高。     

降低受信仪行程开关返程误差方法的分析

由于受信仪的行程开关返程误差较大,导致了运行、射界范围的控制精度不高。 针对返程误差偏大实际问题,研究了受信仪的行程开关工作原理,建立了其工作过程数学模型,采用复合形法开展了参数优化设计。 依据 PDCA 程序,应用因果图分析方法获得影响误差的主要因素,并逐一采取相应的解决措施。 经过验证,其实现了凸轮行程开关的返程误差减小到 0. 5°以下,提高了发射装置运行、射界范围的控制精度,达到了预期的目标。     

PSD法测量大型射电望远镜副面位姿

指向精度是大型射电望远镜天线具有挑战性的关键技术指标.在望远镜运行中,方位俯仰角的变化、重力以及日照等对副面撑杆的综合影响会引起望远镜的副面位姿改变从而引起指向误差的增加和天线效率的降低.基于天马65 m射电望远镜,使用位置传感器装置(PSD)法构建了副面位置偏移量测定系统,可以实时采集副面的3维位移数据,构建重力模型,将测试结果与射电法构建的现有副面模型进行对比,有较好的一致性.此外,该系统可以分析日照(引起撑腿局部温度效应)引起的副面位移情况,也可以监视风载和瞬时启停对副面位姿的影响.