Remote Sensing Image Change Detection with Graph Interaction

Chenglong Liu

2023-07-05

Abstract:Modern remote sensing image change detection has witnessed substantial advancements by harnessing the potent feature extraction capabilities of CNNs and Transforms.Yet,prevailing change detection techniques consistently prioritize extracting semantic features related to significant alterations,overlooking the viability of directly interacting with bitemporal image <a class="link-external link-http" href="http://features.In" rel="external noopener nofollow">this http URL</a> this letter,we propose a bitemporal image graph Interaction network for remote sensing change detection,namely BGINet-CD. More specifically,by leveraging the concept of non-local operations and mapping the features obtained from the backbone network to the graph structure space,we propose a unified self-focus mechanism for bitemporal images.This approach enhances the information coupling between the two temporal images while effectively suppressing task-irrelevant interference,Based on a streamlined backbone architecture,namely ResNet18,our model demonstrates superior performance compared to other state-of-the-art methods (SOTA) on the GZ CD dataset. Moreover,the model exhibits an enhanced trade-off between accuracy and computational efficiency,further improving its overall effectiveness

Computer Vision and Pattern Recognition,Information Retrieval

What problem does this paper attempt to address?

The paper aims to address a key issue in remote sensing image change detection (Change Detection, CD), which is how to identify actual changes between two images of the same geographic area taken at different times. Specifically, existing change detection techniques often focus on extracting semantic features related to significant changes, while neglecting the possibility of directly interacting with bi-temporal image features. To solve this problem, the authors propose a bi-temporal image graph interaction network named BGINet-CD. This method introduces the concept of non-local operations into image feature mapping and proposes a unified self-focusing mechanism to enhance the information coupling between bi-temporal images, effectively suppressing task-irrelevant interference.

Through experimental validation on two public datasets (WHU Building Change Detection Dataset and Guangzhou Dataset), the results show that BGINet-CD not only ensures high accuracy but also achieves computational efficiency optimization, particularly reaching the current State-of-the-Art (SOTA) performance on the Guangzhou dataset. Additionally, ablation experiments further confirm the critical role of the Graph Interaction Module (GIM) in improving model performance. Overall, the method proposed in this paper demonstrates significant advantages in the task of remote sensing image change detection.