基于物联网技术的生物式临震预测关键技术研究

Due to that there is no efficient applicable approach for automatic detection of anomalous animal behaviors and comprehensive analysis is rarely applied to the abnormity detection results from various observation points in biology research related to earthquake, earthquake prediction is less reliable and often failed to be reported or false alarm are declared frequently. The technologies, including intelligent video analysis, cloud computing and the Internet of Things, are introduced into earthquake forecasting in this project. A kind of IOT-based biological monitoring network for impending earthquake prediction is designed and implemented, and four animals, which belong to different categories, are selected as biological indicators of earthquake precursor. With computer vision technology, animals recognition and tracking of the can be performed efficiently, and the postures, actions, activities, active levels, activity frequency of the animals can be detected, and the normal and abnormal behavior of animals can be analyzed and evaluated by considering the factors related to time, season and surroundings to attempt to figure out the correlation of abnormal animal behaviors with earthquake precursors. Based on the information of biological abnormalities from diverse kinds of animals distributed in different geographical area, comprehensive analysis and then the prediction can be made by employing a kind of time-based sequence analysis method and a probability-gain based credibility model, and finally, the forecasting result will be evaluated. As achievements of the research, a kind of biological warning and forecasting mechanism with high credibility for earthquake prediction will be established and the observation will be transferred from focusing on partial animal anomalies to acquiring overall information of abnormal animal behavior and earthquake prediction can be made with high reliability. As a result, a novel approach is offered to learn the relevance of the earthquake precursors with animal abnormities, which are perspective, experiential and random.

针对目前地震生物学缺乏自动检测动物异常行为的有效手段及观测结果的综合分析计算，生物式地震预测存在可靠性低，有漏报和误报等问题。本研究利用智能视频分析、云计算与物联网技术，设计构建基于物联网技术的生物式地震监测网，遴选四种不同种类动物作为地震前兆生物指示器，通过计算机视觉技术对动物进行识别和跟踪，解析动物姿态、动作、行为，并结合时间、季节和环境等元素分析判断动物行为是否异常及其原因，尝试解明动物异常行为与地震前兆的直接相关性，并利用动物异常行为的时间序列分析方法和基于概率增益方法的时空综合分析可信度模型，对分布在不同地理区域内的动物异常信息进行分析和预测并对其进行评估，从而建立一种高可信度的生物式地震预测预警机制，实现从目前动物异常行为的局部性人工观察到全局性高可信度的生物式地震预测的重大转变。为解明趋势性、经验性、不确定性、随机性的动物异常反应与地震前兆的关联问题提供新的探索。

针对目前地震生物学缺乏自动检测动物异常行为的有效手段及观测结果的综合分析计算，生物式地震预测存在可靠性低，有漏报和误报等问题。本研究利用智能视频分析、云计算与物联网技术，设计构建基于物联网技术的生物式地震监测网，遴选四种不同种类动物作为地震前兆生物指示器，通过计算机视觉技术对动物进行识别和跟踪，解析动物姿态、动作、行为，并结合时间、季节和环境等元素分析判断动物行为是否异常及其原因，尝试解明动物异常行为与地震前兆的直接相关性，并利用动物异常行为的时间序列分析方法和基于概率增益方法的时空综合分析可信度模型，对分布在不同地理区域内的动物异常信息进行分析和预测并对其进行评估，从而建立一种高可信度的生物式地震预测预警机制，实现从目前动物异常行为的局部性人工观察到全局性高可信度的生物式地震预测的重大转变。为解明趋势性、经验性、不确定性、随机性的动物异常反应与地震前兆的关联问题提供新的探索。