基于证据理论的可见光和红外融合检测算法  

Fusion Detection Algorithm of Visible Light and Infrared Information Based on Evidence Theory

作　　者：周策 赵秋博 王明杰 宗茂 刘龙[2] ZHOU Ce;ZHAO Qiubo;WANG Mingjie;ZONG Mao;LIU Long(The 54th Research Institute of CETC,Shijiazhuang 050081,China;Xi'an Key Laboratory of Intelligent Spectrum Sensing and Information Fusion,Xidian University,Xi'an 710071,China)

机构地区：[1]中国电子科技集团公司第五十四研究所,河北石家庄050081 [2]西安电子科技大学西安市智能频谱感知与信息融合重点实验室,陕西西安710071

出　　处：《无线电工程》2025年第3期484-492,共9页Radio Engineering

基　　金：国家自然科学基金(62276204);陕西省自然科学基础研究计划(2022JM-336)。

摘　　要：无人机因机动性高和战场适应能力强被广泛应用。然而仅依靠单类传感器的目标检测方法无法满足无人机在高速复杂运动场景下的目标检测要求。针对上述问题,提出一种融合可见光信息和红外信息的融合检测算法,主要包括基于深度学习的图像检测、数据关联和异类信息融合。整合数据关联融合算法并利用证据理论将关联结果进行决策级融合,完成候选目标识别与检测。通过公开数据集OTCBV的部分场景以及无人机航拍数据验证所提方法。实验结果表明,运用多源信息融合检测算法结合多种证据更新YOLO的检测精确率提高至0.9以上。UAVs are widely used for their high mobility and adaptability.However,target detection methods based on a single type of sensors cannot meet the target detection requirements of UAV in high-speed and complex motion scenarios.To address the problems above,a detection and fusion algorithm that integrates visible and infrared information is proposed,which mainly includes image detection based on deep learning,data association and heterogeneous information fusion.The innovation point is integrating data association fusion algorithm and using evidence theory to perform decision-level fusion of association results to complete candidate target recognition and detection.The proposed method is verified by some scenes of public dataset OTCBV and UAV aerial photography data.The experimental results show that the multi-source information fusion detection algorithm updates the detection accuracy rate of YOLO by combining various evidences and improves it to more than 0.9.

关 键 词：无人机 目标检测 DS证据理论 信息融合 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]