自然灾害调查数据的多尺度异常检测方法研究及应用

发布时间：2018-04-15 16:02

本文选题：灾害调查 + 山洪灾害　；参考：《地球信息科学学报》2017年12期

【摘要】：大范围自然灾害调查,涉及区域环境差异大,数据获取方式多样,参与人员多,各级汇总成果中存在一些异常调查单元,需要人工判读其合理性,但单纯依靠人工从海量数据中有效识别异常是不现实的。本文设计了一种自然灾害调查数据的多尺度异常检测方法,综合运用离群检测方法和空间数据挖掘算法,分别进行异常值和异常空间分布模式检测,能够从海量调查数据中快速提取各级尺度的异常值和异常调查单元,支撑人工判读工作。将该方法应用于全国山洪灾害调查评价汇总数据的审核中,以全国历史山洪灾害点和防治区乡镇面积审核为例,分别快速提取了县乡两级区划中的山洪灾害点密度异常单元和面积值异常的乡镇单元,通过对检测结果进行分析,发现是填报口径不一致、单位错误、记录重复等原因造成的。最后分析了该方法在大范围自然灾害调查中的适用条件和方法。
[Abstract]:Large scale natural disaster investigation, involving regional environmental differences, data acquisition methods, many participants, there are some abnormal survey units in the aggregate results at all levels, which need to be manually interpreted its reasonableness.However, it is not realistic to rely solely on manual recognition of anomalies from massive data.In this paper, a multi-scale anomaly detection method for natural disaster survey data is designed. Outliers and spatial distribution patterns are detected by using outlier detection method and spatial data mining algorithm, respectively.It can quickly extract outliers and anomaly investigation units at all levels from massive survey data to support manual interpretation.The method is applied to the audit of the data collected in the investigation and evaluation of the national mountain torrents disaster, taking the national historical mountain torrents disaster spot and the township area audit in the prevention area as an example.The density anomaly unit of mountain flood disaster point and the township unit with abnormal area value in county and township regionalization were quickly extracted. By analyzing the test results, it was found that the results were caused by the inconsistency of filling out caliber, unit error, duplicate record and so on.Finally, the application conditions and methods of this method in natural disaster survey are analyzed.
【作者单位】：天津大学水利工程仿真与安全国家重点实验室;中国水利水电科学研究院;水利部防洪抗旱减灾工程技术研究中心;
【基金】：中国水科院科研专项(JZ0145B042016;JZ0145C022017) 国家自然科学基金项目(51579131)
【分类号】：X43

【相似文献】