Deep Convolutional Neural Network for Mangrove Mapping
Corina Iovan, Michel Kulbicki, Éric Mermet
Abstract
Updated information on the spatial distribution of mangrove forests is of high importance for management plans. Yet, access to mangrove distribution maps is limited, even-though remote sensing data is currently freely available and deep learning algorithms score high performances in automatic classification tasks. The methodologies developed in this paper are based on a deep convolutional neural network and have been tested on WorldView 2 and Sentinel-2 images. The obtained results are highly satisfactory and open perspectives for automatically mapping mangrove distribution over large areas.
Topics & Concepts
MangroveConvolutional neural networkComputer scienceArtificial intelligenceDeep learningRemote sensingArtificial neural networkSpatial analysisPattern recognition (psychology)Distribution (mathematics)Deep neural networksData miningMachine learningGeographyEcologyMathematicsMathematical analysisBiologyCoastal wetland ecosystem dynamicsAgricultural and Environmental ManagementOil Palm Production and Sustainability