SR-Seg
Segmentation of vegetation remote sensing images can minimize the interference of background, thus achieving efficient monitoring and analysis for vegetation information. The segmentation of field phenotypes poses a significant challenge due to the inherently complex environmental conditions. Currently, there is a growing trend of using spectral techniques combined with deep learning for field vegetation segmentation to cope with complex environments. However, two primary limitations persist. On the one hand, the cost of equipment required for real-field spectral data collection is high. On the other hand, the availability of field datasets is limited, and data annotation is time-consuming and labor-intensive. To address these challenges, this study proposes a weakly supervised approach for field vegetation segmentation by using spectral reconstruction (SR) techniques as the foundation and drawing on the theory of vegetation index (VI). Specifically, to reduce the cost of data acquisition, we propose SRCNet and SRANet to reconstruct multispectral images of field vegetation based on convolution and attention modules, respectively. Then, based on the VI, the reconstructed data are feature fused to obtain salient information. Finally, the fused data are segmented in a weakly supervised method, which does not require manual labeling to obtain a field vegetation segmentation map.Our segmentation method can achieve a Mean Intersection over Union (MIoU) of 0.853 on real field datasets, which outperforms the existing methods. In addition, we have open-sourced a dataset of unmanned aerial vehicle (UAV) RGB-multispectral images, comprising 2358 pairs of samples, to improve the richness of remote sensing agricultural data.
SR-Seg main flow diagram.