It is of great significance to reduce the damage caused by flood to quickly and accurately obtain the information of submerged water body and surface features and conduct scientific and effective flood assessment with the help of multi-source remote sensing data.
This study makes use of the characteristics of SAR data that are not affected by cloud and rain weather, and combines the GF-3 pre disaster and disaster radar images with Sentinel-1A disaster radar images through the cooperation of two satellites, giving play to its resolution advantages and time advantages respectively, monitoring and extracting flood water bodies cooperatively, and using threshold segmentation method, object-oriented method and random forest and other methods to achieve the extraction of flood water body information through comparative research, so as to obtain the relevant data sets of flood water bodies. With the help of high resolution optical images GF-3, GF-6, etc., the accurate extraction and statistics of submerged objects are realized, and the submerged object data set is obtained; Finally, through the extracted water body and surface feature information, the risk assessment model and emergency assessment model are constructed to obtain the comprehensive assessment data set of flood disaster. Data results and flood disaster extraction and evaluation algorithms can provide scientific basis for flood disaster prevention and emergency rescue.
1. GF-3 and GF-6 satellite image data: Shandong Liaocheng Data and Application Center of High Resolution Earth Observation System
2. Sentinel-1A and Sentinel-2B satellite image data: European Space Agency（ https://scihub.copernicus.eu/ ）
3. DEM data: Japan ALOS Earth observation satellite 12.5m digital elevation product（ https://earthdata.nasa.gov/ ）
4. Various statistical data of Yinan County: Linyi Statistical Information Network（ http://tjj.linyi.gov.cn/ )
1. Remote sensing data preprocessing:
(1) GF-3 Data Preprocessing
(2) Sentinel-1A data preprocessing
(3) Sentinel 2 data preprocessing
(4) Data preprocessing of Gaofen No.6
2. Extraction of flood water body: Based on ENVI software, three classification methods, namely threshold segmentation, object-oriented and random forest, are used to extract the flood inundation range from SAR images. The overall accuracy and Kappa coefficient are used as indicators to verify the pre disaster water body information extraction results of each classification method. After comparison, we choose to use the extraction results of the object-oriented classification method to determine the flooded area.
3. Extraction of submerged objects: based on ENVI software, three classification methods including object-oriented, random forest and minimum distance supervised classification are used for image classification of land use types. After comparison, we choose to use the classification results of object-oriented classification method to extract submerged features.
4. Establish the evaluation model: firstly, select the factors and normalize them, determine the factor weight by combining the analytic hierarchy process and entropy method, and obtain the risk assessment model and emergency assessment model respectively through the weighted synthesis method. Finally, verify the accuracy.
Data precision:
(1) The overall accuracy of the pre disaster water body information extraction results based on the object-oriented classification method is 95.97%, the Kappa coefficient is 0.9104, and the recall, precision and false alarm rates are 0.9532, 0.9862 and 0.0138 respectively.
(2) The Kappa coefficient of the submerged object extraction result based on the object-oriented classification method is 96.67%.
(3) The accuracy verification of the assessment results is based on Python crawler technology. The proportion of disaster affected points located in areas with higher hazard level and above is 90.32%, and the accuracy of emergency assessment is 92.9%. "
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.ncdc.Hydro.db2453.2022
10.12072/ncdc.Hydro.db2453.2022