基于视频分析的地震检测  

Earthquake Detection Based on Video Analysis

作　　者：黄诚 姚拓中[1] 

机构地区：[1]宁波工程学院电子与信息工程学院,浙江宁波

出　　处：《人工智能与机器人研究》2019年第3期126-139,共14页Artificial Intelligence and Robotics Research

摘　　要：传统的地震烈度分析是基于传统地震测量仪器以及光学式地震仪进行的。虽然这种分析的方法应用广泛,技术相对成熟,但是其响应速度较慢、设备成本高。相比之下,基于视频分析的地震检测是对传统方法的一种补充,有着响应速度快,投入成本较小的优势。因此本文将计算机视觉技术应用到地震分析中,通过运动估计的方法获取视频中的运动矢量序列,由该序列分析得出该视频中是否发生地震,并实现截取地震部分视频、地震的烈度以及绘制其抖动图的功能。再通过灰度值分析、目标分割、特征点匹配等方法滤除一些特殊场景的误判。本文将该算法应用于地震中的现场地震视频和网络上的一些地震视频,并通过实验验证了该算法的有效性。Traditional seismic intensity analysis is based on traditional seismograph and optical seismograph.Although this method is widely used and relatively mature in technology,its response speed is slow and equipment cost is high.In contrast,seismic detection based on video analysis is a complement to traditional methods,which has the advantages of fast response and low cost.In this paper,computer vision technology is applied to seismic analysis.Motion vector sequence in video is obtained by motion estimation method.From this sequence analysis,whether earthquake occurs in the video is obtained,and the function of intercepting seismic video,seismic intensity and drawing acceleration and jitter maps is realized.Then,the gray value analysis,target segmentation,feature point matching and other methods are used to filter out the misjudgments of some special scenes.For the validity of this method,this method is applied to a large number of in-situ seismic videos and some seismic videos on the network.Experiments show that the proposed method can effectively accomplish these tasks.

关 键 词：地震烈度分析 运动目标分割 特征匹配 视频分析 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]