3D Fully Convolutional Neural Networks with Intersection over Union Loss for Crop Mapping from Multi-Temporal Satellite Images (2021-02-15T00:00:00.000000Z)

TL;DR

The capability of a 3D Fully Convolutional Neural Network to map crop types from multi-temporal images and the Intersection Over Union (IOU) loss function for increasing the overlap between the predicted classes and ground reference data are explored.

Abstract

Information on cultivated crops is relevant for a large number of food security studies. Different scientific efforts are dedicated to generate this information from remote sensing images by means of machine learning methods. Unfortunately, these methods do not take account of the spatial-temporal relationships inherent in remote sensing images. In our paper, we explore the capability of a 3D Fully Convolutional Neural Network (FCN) to map crop types from multi-temporal images. In addition, we propose the Intersection Over Union (IOU) loss function for increasing the overlap between the predicted classes and ground reference data. The proposed method was applied to identify soybean and corn from a study area situated in the US corn belt using multi-temporal Landsat images. The study shows that our method outperforms related methods, obtaining a Kappa coefficient of 91.8%. We conclude that using the IOU loss function provides a superior choice to learn individual crop types.

Authors

Sina Mohammadi

2 papers

Mariana Belgiu

1 papers

A. Stein

1 papers

3D Fully Convolutional Neural Networks with Intersection over Union Loss for Crop Mapping from Multi-Temporal Satellite Images

TL;DR

Abstract

Authors

References21 items

Phenology-based sample generation for supervised crop type classification

DeepCropMapping: A multi-temporal deep learning approach with improved spatial generalizability for dynamic corn and soybean mapping

Multi-Scale Interactive Network for Salient Object Detection

CAGNet: Content-Aware Guidance for Salient Object Detection

Attention-Guided Version of 2D UNet for Automatic Brain Tumor Segmentation

Deep learning based multi-temporal crop classification

Temporal Convolutional Neural Network for the Classification of Satellite Image Time Series

Learning a Discriminative Feature Network for Semantic Segmentation

Multi-Temporal Land Cover Classification with Sequential Recurrent Encoders

3D Convolutional Neural Networks for Crop Classification with Multi-Temporal Remote Sensing Images

MODIS phenology-derived, multi-year distribution of conterminous U.S. crop types

Classification of multitemporal SAR images using convolutional neural networks and Markov random fields

A multi-resolution approach to national-scale cultivated area estimation of soybean

Attention is All you Need

Deep Learning Classification of Land Cover and Crop Types Using Remote Sensing Data

DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs

An Assessment of Algorithmic Parameters Affecting Image Classification Accuracy by Random Forests

A meta-analysis of remote sensing research on supervised pixel-based land-cover image classification processes: General guidelines for practitioners and future research

Impact of feature selection on the accuracy and spatial uncertainty of per-field crop classification using Support Vector Machines

“keras

Attention - guided version of 2 d unet for automatic brain tumor segmentation , ” in 2019 9 th International Conference on Computer and Knowledge Engineering ( ICCKE )

Field of Study

Journal Information

Name

Page