Abstract
Infrared small target detection is a research core of infrared image processing and machine vision that is widely applied to civil and military fields. Although the accuracy and efficiency of infrared small target detection have been significantly improved in recent years, the issues of missing detection and false alarms with complex scenes remain because infrared small targets usually have small sizes and weak signals and are submerged by the backgrounds. To cope with the abovementioned problems, we propose in this paper an infrared small target detection method using coordinate attention and feature fusion, named CAFF-Net. First, we present a deep and shallow feature fusion strategy to improve the detection probability, which captures both the low-level structural and textural features and the high-level semantic information of small targets to avoid missing detection. Second, we adopt the coordinate attention mechanism to highlight the target salience and suppress background interference in the feature map, which is able to decrease the false alarm rate with complex scenes. Third, we contribute a real-world infrared small target dataset, named NCHU-SIRST. The constructed dataset consists of 273 training frames and 317 test frames and offers a comprehensive comparison for various infrared small target detection methods with different scenes. Finally, we run our method on the NCHU-SIRST dataset to conduct a comparison with some existing traditional and deep-learning-based infrared small target detection methods. The experimental results demonstrate that the presented CAFF-Net method performs competitively among the comparable methods, especially with the high detection probability and low false alarm ratio under complex scenes. (The code of the proposed CAFF-Net method and the NCHU-SIRST dataset are available at https://github.com/PCwenyue/CAFF-Net-method-and-the-NCHU-SIRST-dataset).
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.