基于混合深度特征的红外目标抗干扰识别算法

Infrared Target Anti-Interference Recognition Algorithm Based on Mixed Depth Features

复杂干扰条件下的红外空中目标识别技术是空战对抗领域的热点研究课题,复杂人工干扰严重遮蔽目标,导致目标特征的连续性与显著性遭到破坏,无法全面描述识别对象的特性,造成空中目标识别准确率下降。针对此问题,提出一种基于图像混合深度特征的空中目标抗干扰识别算法。首先,基于卷积神经网络进行图像深度特征的提取,将深度特征与梯度直方图(Histogram of Gradient,HOG)特征进行有效融合,构建混合深度特征。针对作战场景中的目标与干扰的对抗态势多样性,将支持向量机的二分类模型改进为三分类模型,对目标、干扰以及目标干扰粘连三种状态进行精确分类。实验结果表明:在复杂干扰环境下,基于混合深度特征的空中目标抗干扰识别算法正确率为92.29%,该算法可以有效地解决目标被干扰遮蔽、形成目标干扰粘连状态时的抗干扰识别问题。

The infrared air target recognition technology under complex jamming conditions is a hot research topic in the field of air combat countermeasures.Complex artificial jamming seriously obscures the target,destroying the continuity and significance of the target characteristics.It s unable to fully describe the characteristics of the identified object,resulting in the reduction of the accuracy of air target recognition.In order to solve this problem,an anti-jamming recognition algorithm for aerial targets based on the image mixed depth features is proposed.First,the image depth feature is extracted based on the convolution neural network,and the depth feature is effectively fused with the histogram of gradient(HOG)feature to construct the mixed depth feature.In view of the diversity of the confrontation situation between the target and the jamming in the combat scenario,the two-classification model of the support vector machine is improved to the three-classification model to accurately classify the three states of the target,jamming and target jamming adhesion.The experimental results show that the accuracy of the anti-jamming recognition algorithm based on the mixed depth feature is 92.29%in the complex jamming environment.The algorithm can effectively solve the anti-jamming recognition problem when the target is obscured by jamming and forms the jamming adhesion state of the target.