Search

Crop classification is important for monitoring crop growth and ensuring national food security. Crop plots have a complex spatial planting structure with a certain degree of fragmentation, which leads to mixing of different crops with unclear borders, in turn affecting the ability of the classification model to recognize and distinguish between crops. To address the above problems, this paper proposes a crop classification method based on multi-source optical remote sensing data, which can extract crop information more accurately by taking advantage of the multi-spectral features and different spatial resolutions of different optical data. Firstly, spectral bands and vegetation indexes are extracted from Sentinel-2B and Jilin Gaofen (JLGF02B). Then, a Gemel Convolutional Neural Network (G-CNN) is constructed to fully mine and integrate the feature information at different scales for crop classification. Finally, the method is compared with classic deep learning algorithms. The result shows that the proposed G-CNN algorithm gets the best results with an overall accuracy OA of 96.5% and a Kappa coefficient of 94.7%. The G-CNN model provides a new idea and method for crop classification using multi-scale optical data. [ABSTRACT FROM AUTHOR]

Synthetic Aperture Radar (SAR) images have been widely used for surface water identification due to their all-weather capabilities. However, the presence of inherent speckle noise in SAR data poses a challenge for accurate water identification. Additionally, annotating high-quality water body samples requires significant human labour, which can be costly and time-consuming. Aiming at the above problems, a noise-robust automatic water identification architecture without artificial labels is proposed. First, a two-stage automatic sample collection method that utilizes k-means++ clustering and morphological concepts is designed. Then, a weakly supervised noise-resistant SAR water body segmentation method NRM-ACUNet has been developed based on U-Net combined with LNR-Dice loss function and Conditionally Parameterized Convolutions (CondConv) to minimize the impact of sample noises. Experimental results show that the morphological processing can improve water body sample quality compared to k-means++, and compared with U-Net, NRM-ACUNet performs superior with noise-containing pseudo-samples, achieving 96.8% F1 accuracy and 52.06% accuracy improvement. [ABSTRACT FROM AUTHOR]