[1]易翰枫,卿晓霞,赵劲翔.不透水面数据自动提取的遥感技术方法[J].中国给水排水,2020,36(23):101-107.
YI Han-feng,QING Xiao-xia,ZHAO Jin-xiang.Remote Sensing Technique for Automatic Extraction of Impervious Surface Data[J].China Water & Wastewater,2020,36(23):101-107.
点击复制
YI Han-feng,QING Xiao-xia,ZHAO Jin-xiang.Remote Sensing Technique for Automatic Extraction of Impervious Surface Data[J].China Water & Wastewater,2020,36(23):101-107.
不透水面数据自动提取的遥感技术方法
中国给水排水[ISSN:1000-4062/CN:12-1073/TU]
卷:
第36卷
期数:
2020年第23期
页码:
101-107
栏目:
出版日期:
2020-12-01
- Title:
- Remote Sensing Technique for Automatic Extraction of Impervious Surface Data
- Keywords:
- remote sensing technique; extraction of impervious surface data; rainfall-runoff model; extraction of underlying surface data
- 摘要:
- 高效准确地提取不透水面数据对提高降雨径流模拟效率和精度至关重要。针对中低分遥感（RS）影像存在混合像元、高分RS影像具有光谱局限性的不足，以重庆某小区为研究对象，通过遥感指数法扩充光谱波段，使研究区域易错分地类与其他地类的可分离性得到明显增强，其平均最小分离度由扩充前的1.26、0.16和0.69分别增大到3.29、4.74和8.89，整体平均最小分离度由0.70增大至5.64；基于eCognition Developer平台，采用FNEA算法、K最邻近算法实现了面向对象的不透水面数据自动提取；用混淆矩阵进行精度计算，得到分类总体精度为82.98%、Kappa系数为0.75、不透水面的用户精度为94.85%；研究区域的SWMM模拟结果得到纳什效率系数（NSE）为0.92，径流峰值误差为9.56%，峰现时间误差为0，表明该方法不仅自动化程度高、时效性强，而且模拟精度高，尤其适用于手动提取基本不具备可行性的大尺度区域不透水面数据的提取。
- Abstract:
- Efficient and accurate extraction of impervious surface is crucial to rainfall-runoff simulation. Since the shortcomings of mixed pixels in the low and medium resolution remote sensing (RS) images and the spectral limitations of high resolution RS images, the remote sensing index method was used in a community in Chongqing to extend the spectral band. The separability between easily misclassified land types and other land types in the study area was significantly enhanced. The average minimum separation degree increased from 1.26, 0.16 and 0.69 before expansion to 3.29, 4.74 and 8.89, respectively, and the overall minimum average separation degree increased from 0.70 to 5.64. The object-oriented method based on eCognition Developer with FNEA image segmentation algorithm and K-nearest neighbor classification algorithm was adopted in automatic extraction of the impervious surface data. The results showed that by the confusion matrix the overall accuracy of the underlying surface extraction was 82.98%, the Kappa coefficient was 0.75, and the user accuracy of the impervious surface was 94.85%. The simulation results of SWMM model showed that the peak time error was 0, the Nash-Sutcliffe efficiency coefficient(NSE) was 0.92, and the peak runoff error was 9.56%, which indicated that the method not only had a high degree of automation and strong timeliness, but also had high simulation accuracy. It was especially suitable for efficient automatic extraction of impervious surface data in large-scale area in which manual extraction was not feasible.
更新日期/Last Update:
2020-12-01