地震前兆大数据可视化相似度评价技术研究  

Similarity Evaluation Techniques for Earthquake Precursor Big Data Visualization

作　　者：杨冠泽 单维锋 YANG Guanze;SHAN Weifeng(School of Emergency Management,Institute of Disaster Prevention,Sanhe 065201, China)

机构地区：[1]防灾科技学院应急管理学院,河北三河065201

出　　处：《防灾科技学院学报》2020年第3期42-46,共5页Journal of Institute of Disaster Prevention

基　　金：国家重点研发计划(2018YFC1503806);地震科技星火计划项目(XH16059)。

摘　　要：在web环境中,降采样是解决地震前兆观测大数据可视化中数据显示过于密集和用户交互体验差的一种常用方法。为解决常用降采样算法不能同时保留原数据整体趋势和峰谷细节问题,提出了一种基于最大三角形算法的地震前兆大数据可视化解决方案。为评估此采样算法对基于原始观测数据可视化方案的影响,还提出了一种基于哈希算法的图像相似度评价算法。以17万条地震前兆观测数据为实验数据,对比研究了均值、最大值、最小值、中位数和最大三角形算法(LTTB)等多种采样算法,并使用图像哈希算法对降采样前后的折线图进行相似度评价。实验结果表明,LTTB算法不仅具有较低的时间复杂度,而且能保持原有数据形态和峰谷细节信息,非常适合于地震前兆大数据可视化展示。On the web,as far as the visualization of earthquake precursor observation data is concerned,the down-sampling is a common method to solve the dense data display and the poor user interaction experience.To solve the problem that the common down-sampling algorithm can not keep the overall trend and details of the original data at the same time,we proposed a visualization solution for earthquake precursor big data based on the maximum triangle algorithm.To evaluate the effect of this sampling algorithm on the visualization scheme based on the original observation data,we put forward an image similarity evaluation algorithm based on hash algorithm.Taking 170000 seismic precursory observation data as experimental data,we performed a comparative study on a variety of sampling algorithms such as mean,maximum,minimum,median,and maximum triangle algorithm(LTTB),and then evaluated the similarity of broken line map before and after down-sampling by the image hash algorithm.As the experimental results show,LTTB algorithm not only has low time complexity,but it can keep the shape and peak valley details of original data,which is very suitable for the visualization of earthquake precursors data.

关 键 词：数据可视化 降采样 感知哈希算法 地震前兆数据 LTTB算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]