多源监测数据动态体视化形变模型构建关键技术与应用研究

Various kinds of engineering and natural disasters occur frequently, and it is very necessary to conduct effective monitoring and early warning on the whole life cycle from the surface and the interior of monitoring objects. This project aims at the key technology of dynamic volume visual deformation model construction for multi-source 3D monitoring data. Firstly, in this project, the method of obtaining a high precision multi-source 3D deformation information is studied and an algorithm of automatic processing and deformation extraction for multi-station and multi-period non-continuous time series GB-SAR radar interferogram is proposed; a multi-source monitoring data intelligent matching technology based on artificial swarm intelligence algorithm and special target feature point device is presented. Secondly, the fusion technology of 3D surface deformation model and spatial distribution model is researched, and the dynamic volume visual deformation model based on BIM whole life cycle is constructed. Thirdly, the algorithm of deformation analysis and prediction for POI (points of interest) of the monitoring objects based on time series and finite element analysis is put forward, and the intelligent deformation extraction algorithm based on SVM (support vector machine) for any POI is proposed. Finally, the dynamic volume visual deformation model and analysis system of multi-source 3D monitoring data is developed, and a system of monitoring visualization and analysis and prediction which is suitable for different monitoring objects is built. The research results of this project can improve the precision of disaster monitoring and prediction of natural and engineering structure, and provide surveying and mapping assurance for the strategic management and decision-making of disaster prevention and reduction, which is of potential theoretical and practical value.

各种工程和自然灾害频频发生，如何对其从表面和内部进行全生命周期的高效监测与预警十分必要。本项目针对多源三维监测数据动态体视化形变模型构建技术开展研究。首先，研究高精度多源三维形变信息获取方法，提出适合多站多期非连续时间序列GB-SAR干涉影像自动处理和形变提取算法；提出了基于人工蜂群智能算法和专用目标特征点装置的多源监测数据智能匹配技术；第二，研究监测对象三维表面形变和空间分布模型融合技术，构建基于BIM全生命周期的动态体视化形变模型；第三，提出基于时序分析和有限元分析相结合的监测对象兴趣点变形分析与预报算法，提出基于支持向量机的任意兴趣点形变智能化提取算法；最后，研制多源三维监测数据动态体视化形变模型和分析系统，构建适合不同监测对象的监测可视化与分析预报体系。本课题研究成果不仅可以提高自然和工程结构的灾害监测与预报水平，而且可以为防灾减灾的科学管理与决策提供测绘保障，具有理论和应用价值。

针对自然灾害（如地震、滑坡等）和工程灾害（如溃坝、桥梁垮塌等）威胁着人类生命安全的问题，本项目对多源三维监测数据动态体视化形变模型构建技术开展研究，通过建立自然与工程结构的三维形变模型进行灾害的监测与预警，最大程度减少了灾害带来的损失。首先，针对便携式、轻量级、目标变化快的形变监测需求，团队自主研制了一套小型地基干涉雷达设备Pico-SAR，实现了对形变体的高频率、高精度监测，采用了短轨道多天线移动填充技术，成功地将轨道长度缩短到了30cm，设备的体积、重量和功耗也随之减小，提出了可拆装轨道和异种天线方案以解决设备运输问题，选择高增益发射天线提高信号质量。第二，针对BIM体视化三维重建技术研究，提出了从点云数据采集、语义分割、几何特征提取到BIM建模的BIM全生命周期三维模型构建的技术路线与系统框架，设计了结合注意力机制和全局上下文信息的点云语义分割网络模型获取高精度语义信息，利用RANSAC与最小二乘等算法获取几何结构参数，得到集几何、语义、拓扑的BIM全生命周期体视化模型，为后续室内环境的三维形变监测分析提供高精度的数字化模型。第三，针对形变体的形变分析预报问题，采用支持向量机对各类传感器形变监测信息进行统计分类，并通过人工神经网络算法对形变信息实现智能化提取，采用模糊集理论算法建立多传感器形变信息决策级融合算法获取兴趣点的最佳估值，利用时间序列分析方法与有限元分析方法对形变体变形情况进行准确分析与预报。最后，针对多源监测数据动态体视化建模系统研制与应用研究，自主研发了回转窑测量数据处理系统、地基SAR数据处理系统、近景摄影沉降分析系统、周转型支承架体安装质量快速检测系统等多套系统，通过对SAR、点云、影像等多源数据处理实现对城市基础设施与重大自然灾害的动态三维体视化形变监测预警，为防灾减灾的科学管理与决策支持提供测绘保障，保证了自然和工程结构的安全。