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

作　　者：安浩南 赵明[1,2] 潘胜达[1] 林长青[2] 无[3] AN Hao-nan;ZHAO Ming;PAN Sheng-da;LIN Chang-qing;无(College of Information Engineering,Shanghai Maritime University,Shanghai 201306,China;Key Laboratory of Intelligent Infrared Perception,Chinese Academy of Sciences,Shanghai 200083,China;Shanghai Engineering Research Center of Ship Exhaust Intelligent Monitoring,Shanghai 201306,China)

机构地区：[1]上海海事大学信息工程学院,上海201306 [2]中国科学院智能红外感知重点实验室,上海200083 [3]上海船舶尾气智能监测工程技术研究中心,上海201306

出　　处：《光子学报》2020年第8期170-182,共13页Acta Photonica Sinica

基　　金：中国科学院智能红外感知重点实验室开放课题基金(No.CAS-IIRP-04)。

摘　　要：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.

关 键 词：红外图像 目标检测 伪模态 生成对抗网络 残差网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]