Remote Sensing Water Information Extraction Based on Neural Network Sensitivity Analysis

Characteristics of each image or differences in the surrounding environment will cause the imaging of characteristics cannot keep necessarily balance. So, to avoid the difference caused by using a unified model to extract the water unit, this paper presents a remote sensing water information extraction method based on neural network sensitivity analysis. First, Tchaban and Garson method are utilized to do sensitivity analysis on each band of remote sensing image, and then the bands which are relatively sensitive to the normalized difference water index (NDWI) are obtained. Then ratio computing and threshold segmentation are carried out to extract the water information. Finally, an experimental test on remote sensing water information extraction by using LANDSAT ETM+ remote sensing image data of Long Geer area WanZhongba County in Tibet is conducted. The water information using our method can be extracted accurately.