.png){kind=icon}
.png){kind=icon}
Return
Winter wheat mapping using unbalanced multi-source remote sensing data
.png){kind=icon}
OA
AI
DOI：10.1080/17538947.2025.2509814.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The integration of optical and synthetic aperture radar (SAR) remote sensing images enhances the acquisition of winter wheat planting information and improves the accuracy of winter wheat mapping. However, the distinct imaging mechanisms of multi-source sensors present challenges, particularly when optical remote sensing data suffers from missing information due to atmospheric conditions. This data imbalance can lead to biased optimization during joint learning processes, hindering the network’s ability to capture comprehensive spatio-temporal information on winter wheat. To address this issue, this study introduces a multi-source spatio-temporal feature joint balanced optimization model (MS-JBM). The model enhances winter wheat mapping accuracy by extracting complete spatio-temporal feature information through balanced extraction and complementary two-phase spatio-temporal features. MS-JBM initially leverages similar and stable temporal information from multiple sources to obtain an unbiased representation of multi-source spatio-temporal features. Subsequently, it utilizes complete SAR spatio-temporal features as the primary foundation for constructing winter wheat growth knowledge, guiding the comprehensive interaction of multi-source spatio-temporal features. This process yields complete spatio-temporal feature information of winter wheat, with semantic consistency across multiple interaction rounds ensuring process stability. Comprehensive evaluations on three datasets demonstrate that the proposed MS-JBM model surpasses existing multi-source spatio-temporal feature extraction models.
Keywords:
Winter wheat extraction
multi-source data
deep learning
imbalanced learning
Journal
IF:
4.9
Papers:
1.9K
Citations:
4.7K