Abstract
Convolutional neural network (CNN)-based detection has shown great potential in accurate infrared (IR) ship detection. Typically, IR images exhibit a lack of texture details, whereas the sizes of IR ship targets are extremely multiscale, making it difficult to accurately detect IR ship targets. Herein, we propose a novel strengthened asymmetric receptive field block (SARFB) for accurate IR ship detection. The SARFB contains an asymmetric receptive field block (ARFB), a spatial pyramid pooling (SPP) block, and skip connections. Through these components, SARFB is able to fuse local and global features, enriching the expressive ability and receptive field of the network for multiscale IR ship target detection. Furthermore, because there is no publicly available IR ship target dataset for detection, we created the single-frame IR ship detection (SFISD) dataset, providing the first public benchmark for testing IR ship target detection performance. In comparative studies, the mAP_0.5 of Yolov5 with SARFB reached 93.3%, outperforming other state-of-the-art methods. Finally, we performed experiments on an unmanned surface vehicle (USV) equipped with an IR camera. The results show the superior robustness of our proposed method, especially when target texture information is lacking, and when the IR ship targets are multiscale. The SFISD dataset is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/echoo-sky/SFISD</uri> .
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.