基于生成对抗网络的地质灾害监测异常数据识别方法  

Anomaly Data Identification Method for Geological Disaster Monitoring Based on Generate Adversarial Network

作　　者：刘将成 郝光耀 陶虹 徐岩岩 王亨 江先晖 陈群 LIU Jiangcheng;HAO Guangyao;TAO Hong;XU Yanyan;WANG Heng;JIANG Xianhui;CHEN Qun(School of Software,Northwestern Polytechnic University,Xi’an 710072,China;Shaanxi Institute of Geo-environment Monitoring,Xi’an 710054,China;Key Laboratory of Mine Geological Hazards Mechanism and Control,Ministry of Natural Resources,Xi’an 710068,China;Field Scientific Observation and Research Station of Ground Fractures and Land Subsidence,Shaanxi Province,Ministry of Natural Resources,Xi’an,710054,China)

机构地区：[1]西北工业大学软件学院,西安710072 [2]陕西省地质环境监测总站,西安710054 [3]自然资源部矿山地质灾害成灾机理与防控重点实验室,西安710068 [4]自然资源部陕西西安地裂缝地面沉降野外科学观测研究站,西安710054

出　　处：《高校地质学报》2025年第2期174-184,共11页Geological Journal of China Universities

基　　金：国陕西省重点研发计划项目(2022SF-360);国家自然基金项目(62172335)联合资助。

摘　　要：可靠的地质灾害预警依赖于准确的传感数据。论文针对地质监测传感数据噪声大和长时序特征难以捕捉的等问题,提出一种基于生成对抗网络的地质灾害监测异常数据识别方法。该方法首先引入随机数据增强策略,丰富了训练数据的多样性,提升了模型对噪声的鲁棒性;其次,采用多头自注意力机制提取长时序特征,并通过对抗训练机制提高模型预警性能的稳定性。通过在陕西省地质灾害隐患点提取的4个真实时序传感数据流的实验表明,文章提出的方法在AUROC和F1指标上较现有的机器学习对比方法有5%~10%的提升。Reliable geological hazard warning depends on accurate sensing data.In order to solve the problems of large noise and long time sequence characteristics of geological monitoring sensor data,we propose a method to identify abnormal data of geological disaster monitoring based on generative adversarial network.Firstly,the RandAugment algorithm is used to enrich the diversity of training data and improve the robustness to noise.Secondly,multi-head self-attention mechanism is used to extract long time series features,and the stability of early warning performance is improved by adversarial training mechanism.Experiments on four real-time series sensor data streams extracted from hidden geological disaster points in Shaanxi Province show that the proposed method has a 5%-10%improvement in AUROC and F1 indexes,compared to widely used machine learning methods.

关 键 词：地质灾害监测 传感数据 神经网络 生成对抗网络 异常检测 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]