Back to Search Start Over

Deep-Learning-Based Method for the Identification of Typical Crops Using Dual-Polarimetric Synthetic Aperture Radar and High-Resolution Optical Images

Authors :
Xiaoshuang Ma
Le Li
Yinglei Wu
Source :
Remote Sensing, Vol 17, Iss 1, p 148 (2025)
Publication Year :
2025
Publisher :
MDPI AG, 2025.

Abstract

Timely monitoring of distribution and growth state of crops is crucial for agricultural management. Remote sensing (RS) techniques provide an effective tool to monitor crops. This study proposes a novel approach for the identification of typical crops, including rapeseed and wheat, using multisource remote sensing data and deep learning technology. By adopting an improved DeepLabV3+ network architecture that integrates a feature-enhanced module and an attention module, multiple features from both optical data and synthetic aperture radar (SAR) data are fully mined to take into account the spectral reflectance traits and polarimetric scattering straits of crops. The proposal can effectively address the limitations of using a single data source, alleviating the misclassification problem brought by the spectral similarity of crops in certain bands. Experimental results demonstrate that the proposed crop identification DeepLabV3+ (CI-DeepLabV3+) method outperforms traditional classification methods and the original DeepLabV3+ network, with an overall accuracy and F1 score of 94.54% and 94.55%, respectively. Experimental results also support the conclusion that using multiple features from multi-source data can indeed improve the performance of the network.

Details

Language :
English
ISSN :
20724292
Volume :
17
Issue :
1
Database :
Directory of Open Access Journals
Journal :
Remote Sensing
Publication Type :
Academic Journal
Accession number :
edsdoj.9224fa35688f4d7ab6a44f3f2f7fc6a1
Document Type :
article
Full Text :
https://doi.org/10.3390/rs17010148