杨向敏1，吴 福1，陈柏基1
（1.广西壮族自治区地质环境监测站，广西 南宁 530029）
摘 要：基于物联网技术的广西地质灾害预警预报系统，结合了轻量化模型技术、Spark 空间大数据处理技术、神经网络学习算法等新技术，解决了广西地质灾害监测网络不够完善、智能化水平不高等问题。系统通过3a 运行取得了良好的效果，对于提高突发地质灾害监测预警能力具有重大创新意义。关键词：地质灾害；预警预报；应用模型；物联网；神经网络中图分类号：P208 文献标志码：B 文章编号：1672-4623（2021）03-0112-03
1985年开始，美国在加州旧金山湾、华盛顿州等地区陆续建立了基于降雨监测和雷达监测的区域滑坡和泥石流地质灾害监测预警系统[1]
。美国地质调查局提出的LHP 计划改进了动态滑坡环境的监测技术。欧洲空间组织利用恒定散射体PS 、GPS 以及光学水准测量进行滑坡变形监测。香港特别行政区结合区域滑坡敏感性和临界雨量研究，建立了滑坡和泥石流区域预警系统[2]
。近年来，我国各省相继建立了基于降雨监测的区域地质灾害预警预报系统。
在地质灾害预警预报服务不断深入和发展的新形势下，广西壮族自治区在“广西地质灾害预警预报分析系统”项目成果的基础上，结合市级地质灾害预警预报系统、地质环境综合管理系统、地质灾害综合指挥系统以及基于物联网的地质灾害自动监测预警预报系统，利用新的数据监测与传输技术、数据处理技术和新的地质灾害预警预报方法，开发了新一代广西地质灾害预警预报系统，提升了广西地质灾害预警预报工作的可靠性、稳定性和便利性。目前，该系统已应用于自治区级、市级和县级的区域实时地质灾害监测与预警业务中。
1地质灾害预警预报系统
系统总体结构分为数据采集层、地质灾害数据中
心、地质环境管理、预警预报分析系统和信息发布系统5个部分，如图1所示。根据广西地质灾害的诱发因素和形成机理，系统融合了云计算、大数据、物联网、移动互联网、人工智能等现代化信息技术，通过深度学习算法，建立了基于人工智能的地质灾害动态监测预警神经网络模型[3]
，不断学习、归纳广西地质灾
害的发生规律；通过数据采集系统，收集地质灾害动态监测数据、地下水动态监测数据，接收水文水资源局的实时雨量数据、气象台的实时雨量数据和气象预警信息，构建地质灾害数据中心；通过预警预报分析系统对广西全区进行网格化区域划分，自动选择相应的监测预警模型进行分析，以提高局部区域地质灾害预警预报的精准性；通过信息发布系统为公众提供信息服务。
㻲䶽⯽
▪⯽≷
䰔䛻䃍
⅀䆍
Ⅰ
͙
䉔⣛ ネ⤲<̬ >
图1 系统总体框架图
2地质灾害预警预报技术的主要创新点
2.1 实现地质灾害应用模型的轻量化智能构建
多类型多边界描述的地质灾害预警预报模型，含有大量的因素信息，因此直接建立预报系统较复杂。轻量化智能构建是将空间数据精简为一个点数据，采用GeoHash 编码方式进行编码转换[4]
；再利用细节层次（LOD ）技术
[5]
将海量地质灾害因素信息封装成组
件，集成到地质灾害预警预报系统中，建立应用组件模型库；最后通过应用构建器，以零代码编程的方式
收稿日期：2019-09-03。
项目来源：广西国土资源厅广西地质环境信息系统（一期）资助项目（GXZC2019-G3-1665-GTZB ）。
doi:10.3969/j.issn.1672-4623.2021.03.031
·113·
第19卷第3期快速实现区域地质灾害的统计、分析和管理等多样化的应用需求，如图2所示。
图2 模型轻量化流程图
2.2 基于Spark 的空间大数据技术处理海量地质灾害数据
Spark 是一个大数据分布式编程框架[6]
，将分布式数据抽象为弹性分布式数据集，提供了应用任务调度、RPC 、序列化和压缩的实现方法以及API 上层组件。地质灾害数据主要包括各类专业属性数据、基础地理空间数据、专题空间数据、灾
害点空间数据和其他数据。通过雨量计、水位仪、倾斜仪、智能视频设备等前端专用监测设备对地质灾害进行实时监测；利用物联网技术、互联网技术实现数据的安全接入与管理，建立地质灾害数据中心。在地质灾害预警预报系统中应用基于Spark 的空间大数据技术对海量地质灾害数据进行分析，再通过扩展Spark 实现对大规模地质灾害空间数据的空间查询和快速分析，解决系统分析效率的问题。
2.3 基于神经网络学习算法的智能预警预报系统
神经网络可分为全局逼近和局部逼近两类，其中局部逼近网络具有学习速度快的优点。径向基函数（RBF ）属于局部逼近神经网络[7]
，结构如图3所示。
χ䒿
䮼
䒿
χχ图3 RBF 神经网络
动态自适应RBF 网络模型，无需事先确定隐含层单元的个数，完成聚类所得的RBF 网络是最优的[8]
。RBF 需要选择P 个基函数，每个基函数对应一个训练数据。最常用的RBF 高斯函数为：
R i m
i ()exp 12χ==−−χ2σi C 2
i 2
，,,, 式中，χ为输入向量；C i 为第i 个基函数的中心；m 为感知单元的个数；‖χ-C i ‖为χ与 C i 之间的距离，随
着‖χ-C i ‖的增大，R i (χ)衰减到零。
基于RBF 的插值函数为： F x w X X w X X ()
=−=−+w X X w X X ∑p 22
p
=ϕϕ
1
(p
p
msdnϕϕ(
−++−2
)P )
p
p
22
((1P
))
RBF 神经网络在非线性函数逼近、时间序列分析、模式识别、系统建模、数据分类、图像处理、信息处理等领域得到了广泛应用。
广西地质灾害发生的可能性与灾害所处区域的地质灾害易发性、土壤含水率、累计降雨量、降雨强度以及人类工程活动等因素有关。首先将各区域的影响因素按其对地质灾害发生的贡献进行分级；再将各因素各等级作为输入神经元，发生地质灾害的可能性等级作为输出神经元，通过神经网络进行运算，从而预报地质灾害等级。根据输入信号，通过选取的函数计算输出，并判别输出层的输出是否满足要求，满足则输出，否则，修正权值，继续处理；同时将每次预报对应的实际发生情况作为学习实例对运算进行修正。根据造成网络输出误差因素在输入空间的非均匀分布以及每个RBF 对网络计算贡献的大小，通过策略对参数进行自适应调整，使神经网络的逼近性能、泛化能力达到较高要求。神经网络通过不断改变网络的连接权值，使其输出不断接近期望值。
将神经网络学习算法与广西地质灾害特性相结合，开发学习引擎对诱发地质灾害的相关数据进行深度
学习，挖掘灾害发生与自然、人为因素的关系，获得地质灾害监测预警模型库，并在不断自我学习中对模型进行优化和完善，实现对潜在地质灾害区域的有效预测。
3地质灾害预警预报成果应用
依托信息技术、物联网技术和国土资源“金土工
程”网络建设成果，建立了一个涵盖全区地质灾害数据存储、管理、应用机制的数据资源充分共享的地质
灾害预警预报系统。该系统已在广西自治区级、市级和县级地质灾害预警预报工作中得到了应用，取得了
良好的效果，能高效支撑地质灾害预警预报工作，预防灾害发生、减少损失。广西壮族自治区已在受滑坡、不稳定斜坡威胁的学校安装340套角度倾斜报警仪，在全区的地质灾害易发区中选取了200所学校安装学校地质灾害隐患自动监测站，在全区重要的地质灾害隐患点上安装304套地质灾害隐患自动监测系统，监测数据通过数据采集终端进行实时采集，并传送至广西地质灾害监测预警平台。
杨向敏等：地质灾害预警预报技术创新与应用
地理空间信息
·114·第19卷第3期己二腈
每年的汛期（5～9月）为广西地质灾害的高发期。2016年5月—2018年9月通过该系统进行地质灾害预报和预警信息发布，成功预报地质灾害27起，避免了1 036人的人员伤亡和1 051.5万元的直接经济损失[9]，在广西防灾减灾上取得了良好的效果。
保健食品良好生产规范4结 语
本文介绍了广西地质灾害预警预报系统中轻量化模型技术、Spark空间大数据处理技术、神经网络学习算法等新技术的应用。基于物联网的地质灾害预警预报技术的应用，重要地质灾害点自动监测系统的安装，提升了广西地质灾害自动监测预警能力，提高了广西地质灾害预警预报的智能性、准确性和效率，增强了实时监测和预警预报能力。系统通过3a的运行和预警取得了良好的效果，对于提高突发地质灾害监测预警能力具有重大创新意义。
奥斯卡电影大世界
参考文献
[1] Keefer D K, Wilson R C, Mark R K, et al. Real-time Landslide
三腔二囊管Warning System during Heavy Rainfall[J]. Science,1987, 238(4 829):921-925
[2] Tang C S C, Cheung S P Y. Review of Real-time Data Transmission
Systems for Slope Instrumentation[R]. GEO Report No.262, Hongkong,China,2009
[3] 王继成.一种新的神经网络学习算法[J].计算机工程与科学,
2000,22(3):6-9
[4] 金安，程承旗，宋树华，等.基于Geohash的面数据区域查询[J].
地理与地理信息科学,2013(5):31-35
电工技术杂志
[5] 孟放，查红彬.基于LOD控制与内外存调度的大型三维点
云数据绘制[J].计算机辅助设计与图形学学报,2006(1): 1-8
[6] 胡德敏，龚燕.基于Spark的混合推荐算法研究[J].计算机应
用研究,2017(12):3 585-3 588
[7] 周俊武，孙传尧，王福利.径向基函数(RBF)网络的研究及实
现[J].矿冶,2001(4):71-75
[8] 朱明星，张德龙. RBF网络基函数中心选取算法的研究[J].
安徽大学学报(自然科学版),2000,24(1):72-78
[9] 广西壮族自治区国土资源厅.广西壮族自治区地质环境公报[R].
南宁：广西壮族自治区国土资源厅, 2016-2018
第一作者简介：杨向敏，硕士研究生，高级工程师，主要从事环境地质、地质灾害预警预报信息化工作。
（上接第91页）
还有一种特殊的情况，即线上CPⅡ满足评判标准要求，但CPⅢ控制网复测与建网出现超限的情况，主要表现为坐标较差的变化。分析其原因可能是，CPⅡ点因施工或荷载变化发生了少量变化，但GNSS复测表现为无需更新，采用全站仪进行CPⅢ复测时则表现为坐标较差超限的情况。这种情况虽然不会带来较大的影响，但增加了轨道精调的工作量，因此也建议采用CPⅢ控制网数据对CPⅡ进行更新。
4结 语
线上加密CPⅡ复测评判标准确定的复测成果将作为CPⅢ复测的基准，直接影响轨道精调。本文通过对沪通铁路线上CPⅡ复测数据的分析，对复测评判标准进行了优化，得出应根据不同的轨道类型以及CPⅢ复测的成果分析来确定线上CPⅡ的成果采用情况的结论。由于规范没有进一步说明两次复测成果较差在何种情况下应更新原测坐标成果[8]，因此二次复测如何判定，数据如何使用仍需进行理论推导与实践总结。随着北斗技术的发展与应用，可实时提供精测网最精确的观测成果，线上加密CPⅡ不存在复测的周期[9]也将成为现实。
参考文献
[1] 高速铁路工程测量规范: TB 10601-2009[S].北京：中国铁道
出版社,2009
[2] 铁路工程测量规范: TB 10101-2018[S].北京：中国铁道出版
社有限公司,2018
[3] 邵梽，花向红，赵磊磊，等.高铁控制网复测数据处理与控制
点稳定性分析[J].测绘地理信息,2019,44(6):106-109 [4] 赵建，王新鹏.高速铁路CP I控制网复测成果分析与更新
方法研究[J].重庆理工大学学报(自然科学版),2016,30(2): 133-139
[5] 余鹏.高速铁路运营期线上平面控制网复测方法探讨[J].铁
道勘察,2018(5):12-17
[6] 刘基余. GNSS卫星导航定位的精度、误差与偏差: GNSS导
航定位误差之一[J].数字通信世界,2018(12):1-2
[7] 梅熙，赖鸿斌，李学仕，等.高速铁路工程测量体系建设与创
新[J].高速铁路技术,2020,11(2):68-72
[8] 晏勇，周适.对现行铁路规范中GNSS测量规定的一些探讨[J].
铁道勘察,2020(3):22-27
[9] 王晓凯.高速铁路控制与精密工程测量问题研究[D].西安:
长安大学,2017
作者简介：王作钰，高级工程师，主要从事铁路、城市轨道交通、磁浮交通工程测量等方面的工作。
Index (Vol.19,No.3)
GEOSPATIAL INFORMATION ·VI ·Discussion on the Renewal Principle of GNSS Network in Railway Construction Period by WANG Zuoyu Abstract The online CPII repetition survey of Nantong-Shanghai railway is carried out according to the new specification. However, some points which are renewed according to the evaluation criteria, don’t coordinate with the changes of CPIII network. Through the analysis of several typical cases in the control network repetition survey project, we found that the update of online CPII should be based on the standard evaluation principle, be added to the analysis of CPIII repetition survey and networking results, and be extended to ballastless track. Then, we proposed a method of update CPII with repetition survey data of CPIII network, to maintain the unity of CPIII results and reduce the workload of track fine adjustment.Key words Nantong-Shanghai railway, online CPII, repetition survey, renewal principle (Page:90)Research on Application of Artificial Modeling Technology in 3D Map of Subway by LI Cunwen Abstract The 3D model of the 3D sketch map has a simulation degree of more than 90% and an image element of more than 250 dpi. All kinds of city scenes and information are collected on the spot and drawn according to the relevant standards and requirements of satellite positioning and
national surveying and mapping. The location of the map is precise. The content of the map is not less than 500 meters in the radius of the subway station as the central point, and is overlooked from a 45° angle of view city, so that passengers can easily identify familiar buildings to determine the direction, thereby strengthening the subway exit location and surrounding information guidance function. In this paper, taking 32 stations of Zhengzhou subway line 5 for example, we described the application of artificial modeling technology in Zhengzhou subway transportation service, and studied process and key technology of artificial modeling.
Key words artificial modeling, 3D, 3dsMax, sketch map, baking, subway station (Page:92)Effects of Incidence Angle on the Accuracy of VZ-400 Scanning Point Cloud Measurement by TONG Kui Abstract In the actual measurement of the terrestrial 3D laser scanner, the laser beam reaches the cut-off angle of incidence because of the high scanning target. High incidence angle will lead to the deformation of laser spot, which makes the laser could fall on any part of the light shift, increases the uncertainty, and loses the accuracy. Aiming at VZ-400 scanner, we designed the measurement scheme, and obtained the functions of incidence angle coefficient and d /h by curve fitting, which was convenient to determine the best scanning distance in the measurement scheme.Key words terrestrial 3D laser scanner, incidence angle, accuracy (Page:96)Application of Interpolation and Fitti
ng Method in the Establishment of Elevation Anomaly Model by ZHOU Wei Abstract Different elevation anomaly models have different solutions for normal heights. In order to obtain the optimal model, we generated elevation anomaly surfaces by means of interpolation and fitting method through the uniformly distributed discrete elevation anomaly points in Linyi City. Then, taking the GPS refinement elevation of Linyi City as the detection point, we analyzed the differences between various models.
Key words polynomial interpolation surface, spline function interpolation surface, polynomial fitting, multi-variate regression analysis (Page:99)Application of Star-station Differential GPS System in Offshore Island and Reef Measurement by CUI Jia Abstract The measurement of island and reef plays an important role in China ’s marine land management. In the sea area without the basic control network, the star-station differential GPS system can be used to complete the measurement work efficiently, and its accuracy can meet the requirements of relevant specifications of island and reef measurement. In this paper, we introduced the principle of satellite system, expounded the application of star-station differential GPS system in offshore island and reef measurement. Then, we compared the star-station differential GPS system with the traditional system, and analyzed the advantages of star-station differential GPS system in offshore island and reef measurement. Finally,
we carried out the precision analysis by measurement experiment.Key words star-station differential GPS, island and reef measurement, measurement without control point, positioning technology (Page:102)Spatio-temporal Distribution Characteristic Analysis of Beijing Rail Transit Passenger Flow by CHEN Zhaoning Abstract With the expansion of urban rail transit, the passenger flow continues to increase. Studying passenger travel characteristics is of great significance for the establishment of urban rail transit current-limiting models and the establishment of current-limiting strategies. Based on OD travel data of some urban rail transit line networks in a certain period of Beijing, we studied the spatio-temporal distribution characteristics of passenger travel, and made some suggestions for the Beijing rail transit operating network. The analysis results show that ①during the early rush hours, the travel time of passengers are mainly concentrated at 100～200 min, the travel distances are concentrated at 20～35 km, and the travel distance is linearly related to travel time. ②Most of the departure and arrival points where passengers choose to travel on the subway are concentrated in the Beijing Fifth Ring Road. The passenger flow moves from outside to inside the city. The passenger flows in and out of Chaoyang Distrsict and Haidian District rank in the top two. Key words Beijing City, rail transit, travel characteristic, spatio-temporal distribution (Page:105)Application of Gravity and Magnetic Technology in Iron Ore Survey in Shiling Area, Henan Province by LIU Jiaju Abstract The Shiling working area is located in the southwestern part o
f the Northern Henan Plain, belonging to the residual range of the Funiu Mountains. Combining gravity and magnetic methods in the working area, we carried out the gravity and magnetic inversion by analyzing the distribution, intensity and morphological characteristics of gravity and magnetic anomalies. The result
showed that the anomaly depth of DC8 was close to the horizontal magnetic geological body at about 1.2 km. Then, we verified the magnetic anomaly by drilling. The result showed that it caused by the breccia like andesite of magnetite breccia. Combining with the lithologic and physical properties of the area, we found that magnetite bearing andesite was the important information for searching iron ore deposits, and the main direction of iron ore prospecting.Key words gravity and magnetic inversion, magnetite, andesite, Shiling Area
(Page:109)Innovation and Application of Geological Disaster Early Warning and Forecasting T echnology by YANG Xiangmin Abstract In this paper, we designed an early warning and forecasting system of geological hazards in Guangxi based on the Internet of Things technology. This system combined with the new technologies such as lightweight model technology, Spark spatial big data processing technology and neural network learning algorithm, and solved the problems of imperfect geological hazard monitoring network and low level of intelligence in Guangxi. The system has achie
ved good results through three years of operation, which is of great innovative significance for improving the ability of monitoring and early warning of sudden geological disasters.Key words geological hazard, early warning and forecasting, application model, Internet of Things, neural network (Page:112)Subway T unnel Settlement Prediction Based on GA-SVR by ZHOU Lijun Abstract In this paper, we used support vector regression (SVR )and genetic algorithm (GA )to optimize parameters, and established a subway tunnel settlement prediction model based GA-SVR, which could improve the accuracy of subway tunnel settlement prediction. Firstly, we used the long-term measured subway structure monitoring data to train SVR model, and used GA to optimize the three parameters of the SVR model. Then, we used the method for combining the mean square error of the training model with the leave-one-out cross-validation to determine the fitness of GA. Finally, based on the measured data of tunnel structure settlement of Nanjing subway line 2, we compared and analyzed the predicted value and the measured value. The results show that the predicted value of the subway tunnel settlement predicted by this model is accurate and reliable, and its accuracy can meet the actual requirements of the project.
Key words SVR, GA, deformation prediction, cross-validation, GA-SVR (Page:115)Application of Singular Spectrum Analysis in the Prediction of Subway Settlement by WANG Sijie Abstract Subway
settlement is disturbed by many factors, and its monitoring data often show non-stationary and nonlinear characteristics. In this paper, we used the singular spectrum analysis (SSA )method to extract the trend sequence and period sequence of monitoring data, in order to reduce the noise and improve the signal-to-noise ratio. Then, we used BP neural network to predict and reconstruct the trend sequence and period sequence respectively. Finally, we obtained the predicted value. The experimental results show that the prediction accuracy of SSA_BP neural network model is higher than that of BP neural network model, and the maximum prediction length of SSA_BP neural network model is better.Key words SSA, BP neural network, subway settlement prediction, stability analysis (Page:118)Design and Implementation of Security Information Push System Based on LBS by FAN Zhenlin Abstract With the improvement of positioning accuracy of BeiDou and positioning capability of field terminals, the application of location-based services (LBS )has made new developments in field geological survey. In this paper, we developed a field security information push system based on LBS, supporting the news push of environmental analysis of the current location, the analysis of security measures and the disaster warning service. Then, we discussed and verified the application methods of LBS in the safe production of field geological survey, which could provide a new guarantee support technology for the safe production of field geological survey.
Keywords LBS, safe production, geological survey, information push (Page:121)Quality Control Method of Interior Mapping about Basic Geographical Conditions Monitoring by ZHAO Guihua Abstract Geographical conditions monitoring is a basic work which can fully understand the national conditions, grasp the state affairs and make national policy. Combining with the actual production in 2018 basic geographical conditions monitoring items, we discussed the quality control method of the land cover classification and geographical conditions element results in basic geographical conditions monitoring. According to the typical quality problems in interior mapping, we put forward some countermeasures, which could provide technical support for the regular basic geographical conditions monitoring project.Key words basic geographical conditions monitoring, land cover classification, geographical conditions element, monitoring result, quality control (Page:124)T echnical Method of High Precision Administrative Boundary Surveying and Mapping in Hubei Province by LI Ya Abstract Administrative boundary is an important basis for effective management of the country. With the rapid development of information age, the application demands of multi-scale and high precision administrative boundary data are more and more wide. In this article, we introduced the content and workflow of the 1∶10 000 scale administrative boundary surveying and mapping in Hubei Province based on high-resolution remote sensing images in detail, explored and researched the key technologies of administrative boundary surveying and mapping. Then, we sum
marized and analyzed the typical problems in the administrative boundary surveying and mapping, and provided the corresponding solutions, which could provide experience for the future large scale administrative boundary surveying and mapping.
Key words administrative boundary surveying and mapping, multi-scale, high-resolution, remote sensing image (Page:128)
豫公网安备41160202000603
站长QQ:729038198
关于我们
投诉建议