基于多跳深度网络的红外微小目标检测方法
Infrared Tiny Target Detection Method Based on a Multi-Hop Deep Network
燕舒乐 1陈润宇 1蔡念 1许少秋 1陈健2
作者信息
- 1. 广东工业大学信息工程学院,广东 广州 510006
- 2. 中国科学院长春光学精密机械与物理研究所,吉林 长春 130033
- 折叠
摘要
红外检测作为远程搜索和监视的重要手段,其对微小目标的检测精度影响实际应用价值.为了提高复杂背景下微小目标的测量精度,提出一种基于多跳深度网络的检测框架.首先,为了应对红外微小目标呈现"弱"和"小"的形状特点,利用无锚机制搭建特征金字塔作为骨干网络提取特征图.随后,为了实现渐进式特征交互和自适应的特征融合,在跳跃连接部分设计了由多尺度膨胀卷积组构成的多跳多尺度融合模块.最后,为了降低模型对微小目标位置偏差的敏感性,在训练中使用真实目标与预测目标的Wasserstein距离作为两者的相似性度量.实验结果表明,所提测量方法比已有测量方法具有更好的测量精度和效率.
Abstract
Infrared target detection is an important means of remote search and monitoring,and the accuracy of infrared tiny target detection determines the practical application value of this method.A detection framework based on a multi-hop deep network is proposed to improve the performance of tiny target detection in complex backgrounds.First,to deal with the"weak"and"small"shape characteristics of tiny targets,an anchor-free mechanism is used to build feature pyramids as the backbone for extracting feature maps.Then,to realize progressive feature interaction and adaptive feature fusion,a multi-hop fusion block composed of multi-scale dilation convolution groups is designed at the connection level.Finally,to reduce the sensitivity to position perturbations of tiny targets,the Wasserstein distance between the real and predicted targets is used as a similarity measure.The experimental results show that compared to existing methods,the proposed method delivers better detection performance in terms of accuracy and efficiency.
关键词
红外微小目标检测/多跳深度网络/无锚点机制/Wasserstein距离Key words
infrared tiny target detection/multi-hop deep network/anchor-free/Wasserstein distance引用本文复制引用
出版年
2024
NETL
NSTL
万方数据