基于伪模态转换的红外目标融合检测算法

Infrared Target Detection Algorithm Based on Pseudo Multimodal Images

为提高红外图像目标检测的精度和实时性,提出一种基于伪模态转换的红外目标融合检测算法.首先,利用双循环的生成对抗网络无需训练图像场景匹配的优势,获取红外图像所对应的伪可见光图像;然后,构建残差网络对双模态图像进行特征提取,并采取add叠加方式对特征向量进行融合,利用可见光图像丰富的语义信息来弥补红外图像目标信息的缺失,从而提高检测精度;最后,考虑到目标检测效率问题,采用YOLOv3单阶段检测网络对双模态目标进行三个尺度的预测,并利用逻辑回归模型对目标进行分类.实验结果表明,该算法能够有效地提高目标检测准确率.

In order to improve the accuracy and real-time performance of infrared image target detection,an infrared target fusion detection algorithm based on pseudo modal transformation is proposed.First,the pseudo visible image corresponding to the infrared image is obtained by using the advantage of dual cycle generation confrontation without training image scene matching;then,the residual network is constructed to extract the features of the dual-mode image,and the feature vector is fused by the add superposition method,and the rich semantic information of the visible image is used to make up for the lack of the target information of the infrared image,so as to improve detection accuracy.Finally,considering the target detection efficiency,three scales of dual-mode targets are predicted by using the YOLOv3 single-stage detection network,and the targets are classified by using the logistic expression model.Experimental results show that the algorithm can effectively improve the accuracy of target detection.