A Comparison study of normalized difference water index and object-oriented classification method in river network extraction from landsat-tm imagery

• Published in: 2017 2nd International Conference on Frontiers of Sensors Technologies (ICFST) pp. 198 - 203
• Main Authors: Xike Zhang, Qiuwen Zhang, Gui Zhang, Zifan Gui
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.04.2017
• Subjects: Artificial satellites; Data mining; Earth; Indexes; Landsat-5; NDWI; object-oriented classification method; Remote sensing; river network extraction; Rivers; Vegetation mapping
• DOI: 10.1109/ICFST.2017.8210502

Landsat data have the characteristics of high resolution and wide spectrum, and have been widely used to extract river network. Based on the feature analysis and pretreatment of Landsat-5 Thematic Mapper (TM) data, in this paper, we studied and realized the extraction of river network in Hunan Province, analyzed and dealt with the extraction results of river network. The main research works are as follows: The Landsat-5 TM images were pre-processed by using ENVI5.3 software over the study area in 2011. Radiometric correction, image mosaic and subset were carried out. The Normalized Difference Water Index (NDWI) and object-oriented classification method were used to extract the river network, and then the extraction results of the two methods of residential area, vegetation area and cloud-containing area were analyzed. The results showed that the extraction of river network based on object-oriented classification method was more complete, and the inaccuracy and omission extraction of river network were much less than that based on NDWI. And then, editor tool in ArcGIS was used to delete the non-river network region and connect the non-connected area of the river network.