Comparative study on Extraction of banded water and surface water in urban area based on MNDWI

Abstract

Accurate and rapid extraction of urban water information is the basis of finding out the temporal and distribution of urban water, surveying water pollution, water environmental protection and scientific development and use of the water environment. However, because of the complexity of the background and environment of urban surface water, different water extraction methods have certain differences in extracting urban water. In addition, different water extraction methods face different types of water objects, and their accuracy also has certain differences. Therefore, it is necessary to explore the adaptability of different water extraction methods in different water types in urban areas. Based on the image of landsat8 OLI, this paper selects four sub areas of water in Xi’an, including two sub areas of landsat8 OLI. The Modified Normalized Difference Water Index (MNDWI) is used to extract water from banded and surface water, and compared with the three methods of Normalized Difference Water Index (NDWI), Support Vector Machine (SVM) and Decision Tree Classification. Four precision evaluations are used: the confusion matrix to calculate k, commission errors, omission errors, and overall precision. The advantages of the Modified Normalized Difference Water Index (MNDWI) in extracting banded water and surface water in urban areas are explored. The results show the extraction accuracy of banded water is higher than that of surface water. For banded water, there is little difference between MNDWI and the other three methods in water extraction. For regular surface water, SVM had better extraction effect than MNDWI.MNDWI has the best extraction effect on irregular surface water. In adaptability and stability, MNDWI has the best effect.