Cross spatio-temporal attention network for change detection

Zhigang Yi; Haonan Yu; Zhaoyi Ye; Shengyu Huang; Ying Wang; Liye Mei; Chuan Xu

doi:10.1117/12.3010230

15 November 2023 Cross spatio-temporal attention network for change detection

Zhigang Yi, Haonan Yu, Zhaoyi Ye, Shengyu Huang, Ying Wang, Liye Mei, Chuan Xu

Proceedings Volume 12815, International Conference on Remote Sensing, Mapping, and Geographic Systems (RSMG 2023); 128152U (2023) https://doi.org/10.1117/12.3010230
Event: International Conference on Remote Sensing, Mapping, and Geographic Systems (RSMG 2023), 2023, Kaifeng, China

Abstract

Detecting building changes based on high-resolution remote sensing images constitutes one of the most important tools for urban management, methods based on deep learning are substantially more efficient than traditional methods of detecting changes. Nevertheless, there are several problems with most models, such as incorrect edge detection results and missing detection results. Due to continuous down sampling, there is a loss of detailed information in the image as well as an insufficient ability to characterize features. Our paper proposes a multi-scale feature dependency network to improve the performance of change detection models. By combining the multi-scale embedding module with the spatiotemporal attention module, we can extract the original image features, preserve the rich context with multi-scale convolution, and establish long-short distance feature dependencies across different scale patches using long-short distance attention. Furthermore, we introduce a dynamic position deviation calculation method to accurately determine the relative position relationship between features during the attentional action phase. In this manner, the accuracy of model detection will be improved. Multi-level feature fusion helps improve the model's ability to perceive detail by fusing features at different scales, capturing semantic information from these features, and enhancing the feature representation capability. The publicly available building dataset LEVIR-CD was used to conduct comparative experiments with mainstream methods. In both visual detection and quantitative evaluation metrics, our network outperforms the comparison method.

(2023) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Zhigang Yi, Haonan Yu, Zhaoyi Ye, Shengyu Huang, Ying Wang, Liye Mei, and Chuan Xu "Cross spatio-temporal attention network for change detection", Proc. SPIE 12815, International Conference on Remote Sensing, Mapping, and Geographic Systems (RSMG 2023), 128152U (15 November 2023); https://doi.org/10.1117/12.3010230

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
8 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Feature extraction

Feature fusion

Data modeling

Image processing

Image fusion

Detection and tracking algorithms

Remote sensing

Keywords/Phrases

Search In:

Publication Years