Answers are grounded in this paper and cite their sources.
Change detection (CD) is crucial in tropical rainforest resource management and climate change monitoring research. However, existing methods suffer from inaccurate detection of changing regions for multiscale targets, limited utilization of spatiotemporal information, and inadequate mining of cross-temporal correlations. To address these issues, this article proposes the spatiotemporal difference enhancement network (SDENet) model. The SDENet model solves the problems through a three-step core design: firstly, multilevel feature maps are derived from the Siamese hierarchical adaptive rectangular downsampling; secondly, a novel spatiotemporal difference enhancement fusion module is proposed; finally, a convolutional block attention module Mamba module is designed to capture long-term dependencies. Experiments on the Hainan Tropical Rainforest dataset (RF-CD) demonstrate that SDENet achieves a mean IoU of 0.6879, F1-score of 0.8102, precision of 0.8248, recall of 0.8087, and overall accuracy of 0.8359, outperforming six state-of-the-art methods. On public datasets LEVIR-CD and SYSU-CD, SDENet also achieves competitive results (mIoU 0.7708 and 0.7495, respectively). These results provide effective technical support for rainforest vegetation CD, biodiversity conservation, and ecological restoration assessment.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: