面向单脉冲信号分类的集成特征选择与评价

受大量射频干扰信号影响,快速从海量观测数据中准确识别出单脉冲信号已成为天文数据处理的一项重要任务,而设计和提取有效数据特征,是利用机器学习进行单脉冲信号高效识别的决定因素.针对如何选择最优特征,进而提升单脉冲信号的分类精度这一关键问题,设计了面向单脉冲信号分类的集成特征选择方法.方法首先混合单脉冲信号的参数特征、统计特征和抽象特征,然后分别利用5种单一特征选择方法选出各自的最优特征集,最后利用贪心策略对5种单一方法获取的最优特征集进行集成筛选,获取最优集成特征集.实验表明,最优特征集合既包含统计特征也包含抽象特征.在相同特征数量下,利用集成特征选择比单一特征选择能获得更高的模型精度,可使F1值最高提升1.8%.在海量数据背景下,集成特征选择对减少特征数量、提升分类性能和加快数据处理速度具有重要作用.     

基于卷积神经网络的快速射电暴候选体分类

针对目前从海量的快速射电暴(Fast Radio Burst, FRB)候选体中人工筛选FRB事件难以为继的现状,提出了一种基于卷积神经网络(Convolutional Neural Networks, CNN)的FRB候选体分类方法.首先,通过真实的观测数据和仿真FRB组成训练和测试样本集.其次,建立了二输入的深度卷积神经网络模型,并对其进行训练、测试和优化,获取FRB候选体分类器.最后,利用来自脉冲星的单脉冲数据对该分类器的有效性和性能进行了验证.实验结果表明,该方法可以快速从大量候选体中准确识别出单脉冲事件,极大地提高了FRB候选体的处理速率和效率.     

一种新的基于2维傅里叶谱图像的恒星光谱特征提取方法和深度网络分类应用

天体光谱分类是天文学研究的重要内容之一,其关键是从光谱数据中选择和提取对分类识别最有效的特征构建特征空间.提出一种新的基于2维傅里叶谱图像的特征提取方法,并应用于LAMOST (the Large Sky Area Multi-Object Fiber Spectroscopic Telescope)恒星光谱数据的分类研究中.光谱数据来源于LAMOST Data Release 5(DR5),选取30000条F、 G和K型星光谱数据,利用短时傅里叶变换(Short-Time Fourier Transform, STFT)将1维光谱数据变换成2维傅里叶谱图像,对得到的2维傅里叶谱图像采用深度卷积网络模型进行分类,得到的分类准确率是92.90%.实验结果表明通过对LAMOST恒星光谱数据进行STFT可得到光谱的2维傅里叶谱图像,谱图像构成了新的光谱数据特征和特征空间,新的特征对于光谱数据分类是有效的.此方法是对光谱分类的一种全新尝试,对海量天体光谱的分类和挖掘处理有一定的开创意义.     

基于机器学习的单脉冲搜索候选体识别对FAST观测CRAFTS数据的应用研究

单脉冲搜索作为脉冲星探测的有力工具,在探测旋转射电暂现源以及快速射电暴中扮演着重要角色。为了从海量的射电巡天数据中快速筛选出最有价值的单脉冲搜索候选体,候选体识别已经从早期启发式阈值判断发展到基于机器学习自动识别。对于FAST观测,研究了基于机器学习的单脉冲搜索候选体识别应用到CRAFTS (the commensal radio astronomy FAST survey)超宽带脉冲星数据的性能表现。在评估过程中,使用单脉冲事件组识别(SPEGID)和单脉冲搜索器(SPS)两类自动识别方法,通过7种不同机器学习分类器对CRAFTS基准数据集产生的单脉冲搜索候选体进行自动识别;作为对比,也使用了启发式阈值判断的方法 (RRATtrap和Clusterrank)。结果表明,SPEGID具有最好的性能表现(最高的F1-score值95.1%、次高的召回率95.4%、最低的假阳性率4.7%),SPS具有最快的筛选速度(平均每小时筛选4 010个候选体)。通过对比分析结果,探讨了如何基于FAST观测数据开展高效的单脉冲搜索候选体识别。     

区域GPS网实时计算可降水量的若干问题

目前地基GPS气象学测得的可降水量(PWV)精度好于2mm，但在利用区域GPS网实时计算每个测站上空的PWV时，要涉及到很多常规GPS资料处理时所忽略的问题，如需考虑数据处理软件和计算方式的选择、站坐标的确定和约束、轨道的使用方法、网外辅助站最佳数量的确定、海潮对实时计算PWV的影响以及实时应用于气象服务时的端部效应等问题。利用上海GPS综合应用网获取的2002年6、7月份长江三角洲地区人梅前后的数据，分析了利用区域性的GPS网实时计算高精度的PWV时要解决的各种问题，探讨了其数据处理方案。     

区域GPS网实时计算可降水量的若干问题

目前地基GPS气象学测得的可降水量 (PWV )精度好于 2mm ,但在利用区域GPS网实时计算每个测站上空的PWV时 ,要涉及到很多常规GPS资料处理时所忽略的问题 ,如需考虑数据处理软件和计算方式的选择、站坐标的确定和约束、轨道的使用方法、网外辅助站最佳数量的确定、海潮对实时计算PWV的影响以及实时应用于气象服务时的端部效应等问题。利用上海GPS综合应用网获取的 2 0 0 2年 6、7月份长江三角洲地区入梅前后的数据 ,分析了利用区域性的GPS网实时计算高精度的PWV时要解决的各种问题 ,探讨了其数据处理方案     

射电天文台址常驻电磁干扰检测与识别方法

为有效统计射电天文台址常驻电磁干扰的变化,提高实时电磁环境监测效率,提出一种基于先验信息的常驻电磁干扰信号检测与识别方法.首先通过信噪分离,信号能量估值从电磁环境历史监测数据中提取移动通信,地面数字电视等常驻大信号的中心频率、极化方式、来波方向等特征,经对无线信号传播信道的衰落方式进行分析,提出信号能量模型服从的分布函数假设并采用K-S检验验证该假设,合理设置能量阈值,建立信号模板库.其次,根据信号模板库中的信号特征对实时频谱数据进行双门限能量检测、信号相关性识别,从而提高信号检验的准确率,实现了常驻信号的快速统计.针对射电天文台址实测频谱,运用该方法能够有效识别并统计频谱中的常驻信号,从而为台址干扰缓解策略制定提供重要依据.     

PSR J0835-4510单脉冲的观测研究

利用中国科学院国家授时中心昊平观测站40m射电望远镜, 在L波段对Vela脉冲星(PSR J0835-4510)进行了单个脉冲观测研究. 在56min的观测数据中, 共观测到38040个单脉冲. 探测到观测时间内辐射的所有单脉冲信号, 其中单脉冲的半峰线宽(half-maximum line width, $W_{50     

一种基于MUSIC算法的天地波识别方法

基于我国BPL长波脉冲信号的特征,利用MUSIC(多信号分类)算法对BPL天、地波延迟进行估计,实现天、地波识别。对传统谱估计IFFT(快速傅里叶逆变换)算法和现代谱估计MUSIC算法进行了仿真和比较,结果表明,这两种方法在较低信噪比条件下可有效分离天、地波,且识别误差都能控制在±5μs内,但MUSIC算法比IFFT算法具有更高的精度和分辨率。     

宽带频谱序列干扰信号识别与统计方法

随着科学技术的不断进步,射电天文台站趋于自动化,各类电子设备的广泛使用使得射电天文台站的电磁环境变得尤为复杂,如何有效识别和统计复杂频谱中的干扰信号是当前射电天文台站亟需解决的问题,故提出一种宽带频谱序列干扰信号识别与统计方法.首先,对每组宽带频谱进行信噪分离、识别频谱中的干扰信号;然后,对第1组宽带频谱信号识别结果及信号特征建立模板库,后续每组频谱的信号识别结果与模板库中对应频率的信号进行相似性分析,根据相似性分析结果,统计信号次数,更新模板库;实现宽带频谱序列干扰信号的识别与统计.针对QTT (QiTai Radio Telescope)台站实测频谱,运用该方法进行干扰信号识别与统计,能够有效识别并标记频谱中的干扰信号,并统计干扰信号随时间、方向的变化趋势.     

Random forest algorithm for classification of multiwavelength data

We introduced a decision tree method called Random Forests for multiwavelength data classification. The data were adopted from different databases, including the Sloan Digital Sky Survey (SDSS) Data Release five, USNO, FIRST and ROSAT.We then studied the discrimination of quasars from stars and the classification of quasars,stars and galaxies with the sample from optical and radio bands and with that from optical and X-ray bands. Moreover, feature selection and feature weighting based on Random Forests were investigated. The performances based on different input patterns were compared. The experimental results show that the random forest method is an effective method for astronomical object classification and can be applied to other classification problems faced in astronomy. In addition, Random Forests will show its superiorities due to its own merits, e.g. classification, feature selection, feature weighting as well as outlier detection.     

A New Feature Vector Using Selected Line Spectra for Pulsar Signal Bispectrum Characteristic Analysis and Recognition

Average pulse profiles of pulsar signals are analyzed using the bispectrum tech-nique. The result shows that there are nonlinear phase couplings between the two frequency axes of the bispectrum charts, which indicate nonlinear factors in the generation and prop-agation of pulsar signals. Bispectra can be used as feature vectors of pulsar signals because of their being translation invariant. A one-dimension selected line spectrum algorithm for ex-tracting pulsar signal characteristic is proposed. Compared with selected bispectra, the pro-posed selected line spectra have the maximum interclass separability measurements from the point of view of the whole one-dimension feature vector. Recognition experiments on several pulsar signals received at several frequency bands are carried out. The result shows that the selected line spectrum algorithm is suitable for extracting pulsar signal characteristics and has a good classification performance.     

Automatic Identification and Extraction of Clouds from Astronomical Images Based on Support Vector Machinetwo

For the time-domain astronomical research, the optical telescopes with a small and medium aperture can get a huge amount of data through automatic sky surveying. A certain proportion of automatically acquired data are interfered by clouds, which makes it very difficult to automatically extract the dim objects and make photometry. Therefore, it is necessary to identify and extract clouds from these images as the index figures for a reference in the subsequent information extraction. In this paper, an astronomical image selection system based on the support vector machine is proposed, which sets the gray value inconsistency and texture difference as the reference to select the images interfered by clouds. Based on the classification results, by through the histogram transformation and feature selection, the index figures of clouds can be further extracted. The experimental results show that our method can achieve the real-time selection of astronomical images with a classification accuracy better than 98%. By the histogram transformation and feature selection the index figure of clouds can be preliminarily extracted as the references for the photometry and dim object extraction.     

基于Faster R-CNN的日冕物质抛射检测方法

日冕物质抛射(Coronal Mass Ejection,CME)是一种强烈的太阳爆发现象,对空间天气和人类生活有巨大的影响,因此,日冕物质抛射检测对预报日冕物质抛射、保障人类的生产生活安全具有重要意义。现有的日冕物质抛射检测多采用人为定义特征和界定阈值等方法。由于人为定义特征不能准确表征日冕物质抛射且具有普适性的阈值难于选择,现有的方法对日冕物质抛射的检测效果有待提高。提出一种基于Faster R-CNN(Faster Region-based Convolutional Neural Networks)的日冕物质抛射检测算法。该方法首先结合CDAW(Coordinated Data Analysis Workshop Data Center),SEEDS(Solar Eruptive Even Detection System)和CACTus(Computer Aoded CME Tracking software package)3个著名的日冕物质抛射目录信息,人工标注了包含9113幅日冕图像的数据集,然后根据日冕物质抛射的图像特征较自然图像少、目标尺寸与自然图像有差异等特点,在特征提取和锚点选择方面对Faster R-CNN进行改进。以2007年6月的日冕物质抛射标注数据为测试集,本文算法检出了全部22个强日冕物质抛射事件和151个弱日冕物质抛射事件中的138个,对日冕物质抛射事件的中心角和角宽度等特征参数的检测误差分别在5°和10°以内。     

Research on Star/Galaxy Classification Based on Stacking Ensemble Learning

Machine learning has achieved great success in many areas today, but the forecast effect of machine learning often depends on the specific problem. An ensemble learning forecasts results by combining multiple base classifiers. Therefore, its ability to adapt to various scenarios is strong, and the classification accuracy is high. In response to the low classification accuracy of the darkest source magnitude set of stars/galaxies in the Sloan Digital Sky Survey (SDSS), a star/galaxy classification algorithm based on the stacking ensemble learning is proposed in this paper. The complete photometric data set is obtained from the SDSS Data Release (DR) 7, and divided into the bright source magnitude set, dark source magnitude set, and darkest source magnitude set according to the stellar magnitude. Firstly, the 10-fold nested cross-validation method is used for the darkest source magnitude set, then the Support Vector Machine (SVM), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost) algorithms are used to establish the base-classifier model; the Gradient Boosting Decision Tree (GBDT) is used as the meta-classifier model. Finally, based on the classification accuracy of galaxies and other indicators, the classification results are analyzed and compared with the results obtained by the Function Tree (FT), SVM, RF, GBDT, Stacked Denoising Autoencoders (SDAE), Deep Belief Nets (DBN), and Deep Perception Decision Tree (DPDT) models. The experimental results show that the stacking ensemble learning model has improved the classification accuracy of galaxies in the darkest source magnitude set by nearly 10% compared to the function tree algorithm. Compared with other traditional machine learning algorithm, stronger lifting algorithm, and deep learning algorithm, the stacking ensemble learning model also has different degrees of improvement.     

基于Stacking集成学习的恒星/星系分类研究

机器学习在当今诸多领域已经取得了巨大的成功,但是机器学习的预测效果往往依赖于具体问题.集成学习通过综合多个基分类器来预测结果,因此,其适应各种场景的能力较强,分类准确率较高.基于斯隆数字巡天(Sloan Digital Sky Survey,SDSS)计划恒星/星系中最暗源星等集分类正确率低的问题,提出一种基于Stacking集成学习的恒星/星系分类算法.从SDSS-DR7(SDSS Data Release 7)中获取完整的测光数据集,并根据星等值划分为亮源星等集、暗源星等集和最暗源星等集.仅针对分类较为复杂且困难的最暗源星等集展开分类研究.首先,对最暗源星等集使用10折嵌套交叉验证,然后使用支持向量机(Support Vector Machine,SVM)、随机森林(Random Forest,RF)、XGBoost(eXtreme Gradient Boosting)等算法建立基分类器模型;使用梯度提升树(Gradient Boosting Decision Tree,GBDT)作为元分类器模型.最后,使用基于星系的分类正确率等指标,与功能树(Function Tree,FT)、SVM、RF、GBDT、XGBoost、堆叠降噪自编码(Stacked Denoising AutoEncoders,SDAE)、深度置信网络(Deep Belief Network,DBN)、深度感知决策树(Deep Perception Decision Tree,DPDT)等模型进行分类结果对比分析.实验结果表明,Stacking集成学习模型在最暗源星等集分类中要比FT算法的星系分类正确率提高了将近10%.同其他传统的机器学习算法、较强的提升算法、深度学习算法相比,Stacking集成学习模型也有较大的提升.     

A Comparison of BBN,ADTree and MLP in separating Quasars from Large Survey Catalogues

We compare the performance of Bayesian Belief Networks (BBN), Multilayer Perception (MLP) networks and Alternating Decision Trees (ADtree) on separating quasars from stars with the database from the 2MASS and FIRST survey catalogs. Having a train- ing sample of sources of known object types, the classifiers are trained to separate quasars from stars. By the statistical properties of the sample, the features important for classifica- tion are selected. We compare the classification results with and without feature selection. Experiments show that the results with feature selection are better than those without feature selection. From the high accuracy found, it is concluded that these automated methods are robust and effective for classifying point sources. They may all be applied to large survey projects (e.g. selecting input catalogs) and for other astronomical issues, such as the parame- ter measurement of stars and the redshift estimation of galaxies and quasars.     

Detection of superimposed periodic signals using wavelets

In this paper we present a wavelet-based algorithm that is able to detect superimposed periodic signals in data with low signal-to-noise ratio. In this context, the results given by classical period determination methods depend strongly on the intrinsic characteristics of each periodic signal, like amplitude or profile. It is then difficult to detect the different periods present in the data set. The results given by the wavelet-based method for period determination we present here are independent of the characteristics of the signals.     

Statistical algorithms for identification of astronomical X‐ray sources

Observations of present and future X‐ray telescopes include a large number of ipitous sources of unknown types. They are a rich source of knowledge about X‐ray dominated astronomical objects, their distribution, and their evolution. The large number of these sources does not permit their individual spectroscopical follow‐up and classification. Here we use Chandra Multi‐Wavelength public data to investigate a number of statistical algorithms for classification of X‐ray sources with optical imaging follow‐up. We show that up to statistical uncertainties, each class of X‐ray sources has specific photometric characteristics that can be used for its classification. We assess the relative and absolute performance of classification methods and measured features by comparing the behaviour of physical quantities for statistically classified objects with what is obtained from spectroscopy. We find that among methods we have studied, multi‐dimensional probability distribution is the best for both classifying source type and redshift, but it needs a sufficiently large input (learning) data set. In absence of such data, a mixture of various methods can give a better final result.We discuss some of potential applications of the statistical classification and the enhancement of information obtained in this way. We also assess the effect of classification methods and input data set on the astronomical conclusions such as distribution and properties of X‐ray selected sources. (© 2008 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)