面向地震预警应用的群智感知网络关键技术研究

Accelerometers-equipped smart phones can be manipulated as a “Micro-seismometer”, as they have capability to respond to the earthquakes. As a result, the community sensing network (CSN) which is composed of billions of smart devices becomes the potential Earthquake Early-warning System. The CSN covers the surveillance area with a much more density than the traditional seismic network .As part of mobile internet, CSN can distribute the warning messages to the clients within incredibly short response time. However, the key challenges faced by the accuracy and timeliness of the earthquake early-warning decisions made by CSN, are the large number and the heterogeneous characteristics of the Off-the-Shelf devices. In this project, we investigate the modeling, theory, method and the implementation of the community sensing works whose application purpose is earthquake early warning. First, we will model the characteristics of seismic phase according to the sampling feature of accelerometers on smart phones. Based on the modeling results, the seismic phase detection and timing algorithms are studied. Secondly, we will analyze and improve the current data fusion and filter scheme, design online earthquake warning and situation aware frame. In order to realize the validation and verification of the new model, theory and scheme, we will at last implement a prototype system and all the algorithms based on the mobile internet and cloud. And we will distribute the applications to public mobile device and collect sampling data. The investigation of this project will track out the characteristics of the seismic phase from the samplings of common accelerometers. This project will cross the barriers between the community sensing network and the earthquake early warning system, and provide the reference for other applications of the community sensing network.

能对地震震动进行响应的普通智能手机加速度传感器，犹如一台“超微型测震仪”。而由无数台智能手机等设备组成的群智感知网络，则成为潜在的地震预警系统，不仅具备远大于传统地震台网的覆盖密度，还能利用移动互联网极快的信息同步速度，实现预警信息向海量终端的实时发布。然而，群智感知网络中非定制传感器的海量规模和异构特征给地震预警的准确性和时效性提出了挑战。本项目针对地震预警应用的群智感知网络，从模型、理论与方法以及系统实现三个方面展开研究。首先，根据加速度传感器对地震事件的响应特性，研究非定制器件采样的震相特征模型；在此基础上，探索移动终端地震检测和事件时间测定的理论与方法，利用数据融合和动态滤波机制研究实时连续地震预警和信息同步策略；进而设计开发基于移动互联网的实验预警系统，分析验证模型、理论和方法的有效性。通过本项目的研究，将为群智感知网络在地震预警应用及其他各领域的发展提供理论和实践上的参考依据。

项目针对普通智能手机对地震预警的应用，从模型、理论和方法以及系统体系结构等领域展开研究。具体的研究从数据采样和分析开始，重点研究了智能手机对地震等异常事件的响应特征和异常事件检测算法以及相应的数据融合算法等。取得了以下六个方面的进展：1.基于通用智能手机采样特征，设计实验场景进行数据采样和波形特征分析，研究了智能手机采样数据的降噪算法，建立了地震震相特征模型。2.针对不同的智能手机系统，分别对常用的地震震相捡拾算法进行测试，找出了通用震相捡拾算法在手机上的移植方法。3.根据智能手机采样中的精度缺陷问题和噪声问题，设计了加速度计和陀螺仪融合校准的数据处理机制。4.设计了基于动态时间弯曲距离的双门限阈值异常事件检测算法，提高了系统的检测准确率。5.针对单一智能手机采样精度低下问题，设计了海量智能手机数据融合的交互融合机制并理论分析了其误警率性能。6.利用实验室的小型结构振动台等器件，模拟采样了智能手机的地震检测数据，并对上述提出的算法和机制进行了验证和优化。.存在的相关问题是：1.目前的智能手机对真实地震事件的采样数据极度缺乏，也导致地震震中分析算法和地震烈度测定算法难以实施2.同样由于实验规模的限制，海量手机的数据交互机制和信息洪泛机制仍然没有没有取得进展。这两个方向的研究目前正在继续，项目组将扩大实验范围，长期进行相关研究工作。项目的研究中已经发表论文5篇，申请专利2项，培养硕士研究6人。