| 摘要: |
| 为应对现代战争中隐身飞行器、小型无人机及高速导弹等微弱空中目标检测的技术挑战，提出了一种基于雷达与红外图像融合的智能微弱目标检测算法。该方法首先对多帧红外图像进行最大值累积，以增强目标航迹的空-时相关性特征；随后利用二坐标雷达的测量数据，将潜在目标位置投影至红外图像中，从而提高目标航迹在红外图像中的连贯性。在检测阶段，将Dimension-Aware Selective Integration(DASI) 模块和Large Separable Kernel Attention(LSKA) 模块引入YOLOv8实例分割网络，优化网络结构，以实现对微弱目标的精确检测。试验结果显示，在信噪比为8 dB的条件下，该算法对匀速直线运动、匀加速运动及复杂机动目标的检测概率均超过80%。 |
| 关键词: 微弱目标检测 雷达-红外图像 信息融合 |
| DOI：10.20079/j.issn.1001-893x.241127005 |
|
| 基金项目:国防科技重点实验室基金项目（2023-JCJQ-LB-016） |
|
| Radar-Infrared Image Fusion-based Intelligent Weak Target Detection Algorithm |
| SUN Dianxing,ZHANG Yihong,PENG Ruihui |
| (Qingdao Innovation and Development Base，Harbin Engineering University，Qingdao 266000，China) |
| Abstract: |
| In response to the technical challenges of detecting weak aerial targets such as stealth aircraft，small unmanned aerial vehicles，and high-speed missiles in modern warfare，an intelligent weak target detection algorithm based on radar and infrared image fusion is proposed.The method first accumulates the maximum values of multiple infrared frames to enhance the spatio-temporal correlation features of the target trajectory.Then，through using the measurement data from a bi-coordinate radar，the potential target locations are projected onto the infrared images，improving the coherence of the target trajectory in the infrared imagery.During the detection phase，the Dimension-Aware Selective Integration(DASI) module and Large Separable Kernel Attention(LSKA) module are integrated into the YOLOv8 instance segmentation network to optimize the network structure for precise detection of weak targets.Experimental results show that under a signal-to-noise ratio of 8 dB，the detection probability for targets with uniform straight-line motion，uniform acceleration，and complex maneuvers exceeds 80%. |
| Key words: weak target detection radar-infrared image information fusion |
《电讯技术》编辑部 