A novel surface water index using local background information for long term and large-scale Landsat images

摘要: Abstract Surface water plays a vital role in natural environment and human development. The research of extraction method using remote sensing image is hot topic, which has been widely developed index, classification, subpixel, other aspects. Compared with methods, water-index based the advantages fast speed convenience. characteristics surface water, such as wide coverage instability, make index stand out monitoring large area water. However, land complex, main factors that reduce accuracy are also different, shadow urban areas leakage unshaded areas. current bound to weaken information body when suppressing shadows, vice versa. To address these issues, contrast difference (CDWI) (SDWI) proposed this paper by improving modified normalized (MNDWI). CDWI used enhance information, suitable for without building shadows. SDWI eliminate buildings, Moreover, background (BDWI) was combining through regularizer B, extract under complex background. B represents similarity between local features reference area, locally weight CDWI, so BDWI can automatically shadowless buildings. results BDWI, MNDWI, tasseled cap wetness (TCW), automatic (AWEInsh, AWEIsh), 2015 (WI2015) were comparison. Other methods tend perform well only built-up or non-built-up areas, while various backgrounds high stability. overall produced 91.58–97.57%, 84.85–97.09%, 81.63–94.40%, MNDWI 80.19–95.64%, TCW 82.33–95.98%, AWEIsh 87.50–96.37%, AWEInsh 80.59–98.78%, WI2015 78.24–98.38%. Combining helpful improve environment. Finally, Jiangsu Province, China extracted changes 1985, 2000, analyzed.