Multilayer Perceptron Neural Network for Surface Water Extraction in Landsat 8 OLI Satellite Images

摘要: Surface water mapping is essential for monitoring climate change, resources, ecosystem services and the hydrological cycle. In this study, we adopt a multilayer perceptron (MLP) neural network to identify surface in Landsat 8 satellite images. To evaluate performance of proposed method when extracting water, eight images typical regions are collected, index support vector machine employed comparison. Through visual inspection quantitative index, algorithm terms entire scene classification, various types noise suppression comprehensively compared with those machine. Moreover, band optimization, image preprocessing training sample analyzed discussed. We find that (1) based on evaluation, extraction using MLP better than or The overall accuracy ranges from 98.25–100%, kappa coefficients range 0.965–1. (2) can precisely extract effectively suppress caused by shadows ice/snow. (3) 1–7-band composite provides optimization strategy algorithm, high-quality samples benefit classification. future studies, automation universality be further enhanced generation newly-released global products. Therefore, has potential map series other high-resolution implemented mapping, which will help us understand our changing planet.