We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
EEA-Net: edge-enhanced assistance network for infrared small target detection.
- Authors
Wang, Chen; Hu, Xiaopeng; Gao, Xiang; Wei, Haoyu; Tao, Jiawei; Wang, Fan
- Abstract
With the development of deep learning, the performance of infrared small target detection (IRSTD) has been significantly improved. A precise shape of the target edge is crucial for segmenting small infrared targets. However, existing CNN-based methods do not effectively integrate the edge and shape information of the target, leading to the distortion of the detected target shape. To alleviate this problem, this paper proposes an edge-enhanced assistance network (EEA-Net) for IRSTD. Specifically, we design an edge-gate module (E-G) to extract the edge information and convey it to deep layers through a cascading structure. Based on E-G, we further propose an information-preserving attention module (IPA) to reduce the loss of edge information in the network. In addition, we design a global perception module (GP) to facilitate the flow of edge information and enhance the network’s ability to fuse information at multiple scales. Extensive experiments on the public datasets demonstrate the superiority of the proposed method over state-of-the-art IRSTD methods.
- Publication
Machine Vision & Applications, 2024, Vol 35, Issue 4, p1
- ISSN
0932-8092
- Publication type
Article
- DOI
10.1007/s00138-024-01554-y