人工智能
变压器
计算机科学
特征(语言学)
计算机视觉
电气工程
电压
工程类
语言学
哲学
作者
Chuanqin Li,Zhanchao Huang,Xiaoming Xie,Wei Li
标识
DOI:10.1016/j.infrared.2023.104723
摘要
Infrared small target detection (ISTD) technology plays an important role in air defense early warning systems, earth observation and military reconnaissance. However, ISTD faces challenges, such as small target size, complex background, and interference of local flash or blind elements, which constrain the performance improvement of existing methods. In this regard, an infrared small target transformer network (IST-TransNet) is proposed. In the proposed IST-TransNet, an anti-aliasing contextual feature fusion (ACFF) module is designed to improve the translation invariance during contextual feature fusion for locating infrared small targets accurately. To suppress complex background interference and highlight the features of infrared small target, a spatial and channel attention (SCA) module is developed. Furthermore, a vision-transformer branch (VTB) is proposed to eliminate the interference of local flash elements by introducing non-local correlation features. Extensive experiments on the public dataset demonstrate that the proposed IST-TransNet can effectively improves the performance of ISTD by maintaining translation invariance of contextual features and suppress the interference of background and local flash elements.
科研通智能强力驱动
Strongly Powered by AbleSci AI