This is the authors' abstract. We don't add key points for this paper.
Change detection in remote sensing imagery is a crucial technique for Earth observation, primarily focusing on pixel-level segmentation of change regions between bi-temporal images. The core of pixel-level change detection lies in determining whether corresponding pixels in bi-temporal images have undergone changes. In deep learning, the spatial and channel dimensions of feature maps represent distinct types of information derived from the original images. In this study, we discovered that in change detection tasks, difference information can be computed not only from the spatial dimension of bi-temporal features but also from the channel dimension. Based on this insight, we designed the Channel-Spatial Difference Weighting (CSDW) module, which serves as an aggregation-distribution mechanism for bi-temporal features in change detection. This module enhances the sensitivity of the change detection model to difference features, thereby improving its overall performance. Furthermore, bi-temporal images share the same geographic location and exhibit strong inter-image correlations. To effectively capture and utilize these correlations, we designed a decoding structure based on the Layer-Exchange (LE) method, which enhances the interaction of bi-temporal features. This approach allows the model to better leverage the temporal dependencies between the images, leading to more accurate change detection. To validate the effectiveness of our proposed method, we conducted comprehensive experiments on four widely used datasets: CLCD, PX-CLCD, LEVIR-CD, and S2Looking. The experimental results demonstrate that the proposed LENet model significantly outperforms existing methods in terms of change detection accuracy. The code and pre-trained models have been publicly available at: https://github.com/dyzy41/lenet, facilitating further research and application in the field.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Dong et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: