高分辨率D-TomoSAR组合形变模型与P-TWIST方法研究

With the rapid development of major cities and dramatic increase of city load in China, structural degradation of buildings and collapse of elevated bridges have occurred in many cities, resulting in great property damage. Therefore, continuous and precise monitoring of urban man-made structures is of great importance to local safety. Time series InSAR, although has been demonstrated as an advanced remote sensing technology for deformation monitoring in many cases, suffers from low accuracy in layover areas caused by side-looking geometry of SAR instruments. Aiming at solving the layover problem and realizing precise deformation monitoring, this project would carry out research on high-resolution D-TomoSAR models and innovative methodology, which would be composed of three work packages... In TomoSAR system model, the back-scattering signal is only sampled along elevation direction. However, another sampling along velocity direction is also necessary in D-TomoSAR, leading to dramatic increase in the grant measurement matrix. Targeting at this problem, a novel tomographic approach named pooling based two-step iterative shrinkage thresholding (P-TWIST) is proposed in this project... In current D-TomoSAR system model, linear deformation model is assumed for scatterers, which is not always correct for man-made structures. An integrated deformation model which considers both linear and seasonal deformation of scatterers would be implemented, replacing the simple linear deformation model.. . Besides linear and seasonal deformation model, high-rise buildings and bridges also suffer from high frequency oscillation problem, which would severely destruct the D-TomoSAR reconstruction accuracy. Another key technology is to simulate the high-frequency oscillation with stochastic deformation model and integrate it into the combined deformation model. With the integrated deformation model, a more precise P-TWIST based D-TomoSAR analysis would then be carried out, estimating the linear deformation, seasonal deformation and oscillation at the same time... The outcome of this research project will overcome the limitation of current high-resolution D-TomoSAR methodology in precise deformation monitoring of man-made structures, and eventually provide service for the government in disaster prevention and reduction.

随着我国城市化进程的推进及城市荷载的增加，导致了“楼歪歪”、“楼脆脆”、高架桥坍塌等事故的发生，引起了社会对人工地物精细形变监测的高度重视。时间序列InSAR技术由于假设像元内只存在单一主导散射体，在“叠掩”区域形变监测精度受限。本项目针对这一问题，开展高分辨率D-TomoSAR模型与方法研究。具体研究内容可以分为：针对由TomoSAR拓展到D-TomoSAR产生的测量矩阵维度激增问题，提出了基于卷积特征池化的双步迭代收缩阈值(P-TWIST)关键技术;针对现有D-TomoSAR线性形变模型过于简单问题，研究更加合理的基于线性和季节性两种形变的组合形变模型；针对部分人工地物的高频摆动问题，为高频摆动建立随机形变模型并纳入组合形变模型当中，可同时估计线性、季节性形变和摆动误差，提高形变监测精度。本项目的研究成果将推动D-TomoSAR技术的应用水平，为政府防灾减灾提供服务。

随着我国城市化进程的不断推进及城市荷载的增加，导致了诸如城市沉降、地陷、“楼歪歪”、“楼脆脆”、高架桥坍塌等事故的发生，引起了社会各界对城市人工地物精细形变监测工作的高度重视。单一轨道的时间序列InSAR系统模型由于假设像元内只存在单一主导散射体，在“叠掩”区域形变监测精度收到限制。本项目针对这一问题，开展高分辨率D-TomoSAR模型与融合时序SAR数据融合方法研究，具体研究内容可以分为：.（1）针对由TomoSAR拓展到D-TomoSAR产生的测量矩阵维度激增问题，提出了基于卷积特征池化的双步迭代收缩阈值(P-TWIST)关键技术，基于最小冗余线阵理论，实现了D-TomoSAR系数基线阵列优化设计，实现了基于线性形变模型的建筑物差分层析重建，提取了精细的三维结构及变形信息，成功应用于沈阳市主城区高精度变形监测。.（2）针对多传感器SAR数据存在的监测时段不连续、升降轨数据无法共用、难以适用高精度、长时间形变监测等问题，提出了两种时序SAR数据融合方法。针对存在时间间隔的多传感器SAR数据，提出了一种基于岩土离心模型的融合方法，并应用于沈阳市浑河沿岸长时间序列监测。针对存在时间重叠的升降轨SAR数据，提出了一种基于Kriging插值和奇异值分解的融合方法，提取了沈阳市铁西区的高精度变形监测结果。.（3）针对露天矿边坡等特殊场景，提出了一种基于地形起伏的三维变形估计方法，采用激光雷达生成的高精度DEM数据，求解了露天矿采坑边坡、排土场及尾矿库坝的坡向精细变形，并与地面测量数据进行了交叉对比验证，成功应用于鞍钢集团鞍千矿及大孤山铁矿高精度变形监测。.总体而言，本项目的研究成果推进了层析SAR及时序SAR形变监测方法的改进，促进了SAR遥感监测精度的提升，有利于帮助相关人员掌握人工地物精细形变信息，为政府防灾减灾及安全监测部门提供服务。.

内容获取失败，请点击重试