协同SAR与光学遥感的城市地下供水管网漏点探测研究

Research on Synergistic SAR and Optical Remote Sensing for Leak Detection in Urban Underground Water Supply Pipeline Networks

摘要: 针对城市地下供水管网漏损识别中传统地面检测效率受限、区域覆盖不足等问题，本文以北京五环内城区为研究区，联合Sentinel-1雷达与Sentinel-2光学遥感影像，针对植被覆盖条件下的反演土壤含水量及地下水管漏点探测进行研究。首先，基于AIEM模型定量分析裸土后向散射系数与土壤含水量及雷达入射角的关系，构建土壤含水量估计模型；其次，引入水云模型并结合NDWI估算植被含水量，构建植被覆盖条件下的表层土壤含水量半经验反演模型；在此基础上，利用PCA与SVM相结合的方法对疑似漏点进行识别与分类。结果表明：基于3景Sentinel-1 SAR影像、3景Sentinel-2光学影像及738个有效样本建立的模型能够较好表征研究区土壤水分空间分异特征，其中VV极化建模效果优于VH极化；模型内部验证相关系数为0.6398，均方根误差为0.0677。在疑似漏点POI中，共有64个POI成功检出漏点，预测精确率达到74.4%。本研究论证了基于SAR与光学遥感协同的土壤含水量反演方法能够为城市地下供水管网漏损识别提供有效的技术支撑，对城区地下管线的大范围快速筛查具有一定的工程应用价值。

Abstract: To address the limitations of conventional ground-based approaches for urban underground water supply pipeline leakage detection, particularly their low efficiency and insufficient regional coverage, this study selected the urban area within Beijing’s fifth ring road as the study area and integrated Sentinel-1 data with Sentinel-2 optical remote sensing data to investigate surface soil moisture inversion under vegetation cover and leak detection in underground water pipelines. First, the AIEM model was employed to analyze the response relationships among bare-soil backscattering coefficients, soil moisture, and radar incidence angle, thereby establishing a soil moisture inversion model for bare-soil conditions. Then, a semi-empirical inversion model for surface soil moisture under vegetation cover was developed by incorporating the water-cloud model and estimating vegetation water content using the normalized difference water index (NDWI). On this basis, a combined PCA-SVM approach was applied to identify and classify suspected leakage points. The results show that the model, developed using three Sentinel-1 SAR scenes, three Sentinel-2 optical scenes, and 738 valid samples, can effectively characterize the spatial variability of soil moisture in the study area, with VV polarization outperforming VH polarization in model construction. Internal validation yielded a correlation coefficient of 0.6398 and a root mean square error of 0.0677. Among suspected point-of-interest (POI) locations, leakage was successfully detected in 64 POIs, corresponding to a prediction accuracy of 74.4%. These findings demonstrate that the synergistic use of SAR and optical remote sensing for soil moisture inversion provides effective technical support for leakage identification in urban underground water supply pipeline networks and shows certain engineering application value for large-scale and rapid screening of urban subsurface pipeline systems.