基于改进YOLOv7的机载红外弱小目标检测算法

Airborne infrared dim target detection algorithm based on improved YOLOv7

随着现代化战争的技术升级,机载红外探测领域对更快更远更准地发现目标的需求日益强烈。为满足机载环境下对红外弱小目标高精度高帧率的检测,本文提出了一种基于YOLOv7改进的目标检测算法,以YOLOv7目标检测算法为基础,进行了修改网络结构和加深卷积层数来使特征提取更多的小目标信息特征;并对骨干网络获取的特征层引入注意力机制来提高神经网络对小目标的感知能力以及提高小目标所在区域的权重占比;使用EIOU损失函数替换原本的CIOU损失函数,提高了收敛速度和定位精度。实验结果表明,相较于原算法YOLOv7,在极小损失帧率的情况下,改进后的算法mAP可以达到98.49%,相较原始算法提升了1.24%,有助于提升对机载红外弱小目标的检测准确率。

With the technological upgrades of modern warfare,there is a growing need for faster,farther and more accurate target detection in the field of airborne infrared detection.In this paper,an improved target detection algorithm based on YOLOv7 is proposed to meet the high-precision and high-frame-rate detection of infrared dim dim targets in airborne environment.Firstly,based on the YOLOv7 target detection algorithm,the network structure is modified and the number of convolutional layers is deepened to extract more features of small target information.Moreover,the attention mechanism is introduced into the feature layer obtained by the backbone network to improve the perception ability of the neural network to perceive the small targets and increase the weight share of the region where the small targets are located.Finally,the EIOU loss function is used to replace the CIOU loss function,which improves the convergence speed and positioning accuracy.The experimental results show that compared with the original algorithm YOLOv7,the improved algorithm can reach 98.49%mAP with minimal loss of frame rate,which is 1.24%higher than the original algorithm,and it helps to improve the detection accuracy of airborne infrared dim small targets.