WATER BODY INFORMATION EXTRACTION FROM HIGH RESOLUTION SENTINEL-1 IWS MODE SAR IMAGES USING LI’S MINIMUM CROSS ENTROPY THRESHOLD METHOD
ISBN 978-85-88783-11-9
Authors
1Nguyen, D.; 2Tran, G.; 3Nghiem, H.
1HANOI UNIVERSITY OF MINING AND GEOLOGY, VIETNAM Email: nguyenbaduy@humg.edu.vn
2HANOI UNIVERSITY OF MINING AND GEOLOGY Email: tranthihuonggiang@humg.edu.vn
3BAC GIANG AGRICULTURE AND FORESTRY UNIVERSITY Email: nghiemthihoai@gmail.com
Abstract
As stated in the fourth World Water Development Report (WWAP, 2012): “Information about water supply and use is becoming increasingly important to national governments, who need reliable and objective information about the state of water resources, their use and management. Farmers, urban planners, drinking water and wastewater utilities, the disaster management community, business and industry, and environmentalists all need to be informed”. SAR has the advantage of all-time all-whether for the earth observation, it is not affected by night and fog and which has the penetrating ability for the cloud interference during the flood period. The new generation of SAR satellites (has a much improved spatial, temporal and radiometric resolution), Sentinel-1A launched on 3 April 2014 (completed commissioning on 23 September 2014 with four exclusive imaging modes of operations: Interferometric Wide Swath (IW), Extra Wide Swath (EW), Strip Map (SM) and Wave (WV). Due to the low backscattering coefficient of water body in the SAR images, the water and land boundary is obvious, image segmentation - the threshold methods are often used in water body extraction from SAR images. In this study, we present an automatic method to identify water body areas by applied the method of minimum cross entropy also known as Li's minimum cross entropy method (chooses "the best threshold" which loses less information during the thresholding) from Sentinel-1 IW images. The results of proposal method compared to the reference data indicate the Completeness (User accuracy), Correctness (Producer accuracy) and Quality (Overall accuracies) of 99.5%, 98.8 % and 97.9 % respectively. The method is straightforward, easy to implement and might be applied for other areas even on a regional or global scale. In addition, it improves the automatic identification level of computer water body identification to a large extent, promoting the progress of flood disaster research.