基于知识图谱的认知语义通信目标检测系统  

Cognitive semantic communication object detection system based on knowledge graph

作　　者：刘伯阳 黄一杰 孙连锐 李泽 赵云 LIU Boyang;HUANG Yijie;SUN Lianrui;LI Ze;ZHAO Yun(School of Communications and Information Engineering,Xi’an University of Posts and Telecommunications,Xi’an 710121,China;Shaanxi Key Laboratory of Information Communication Network and Security,Xi’an 710121,China)

机构地区：[1]西安邮电大学通信与信息工程学院,陕西西安710121 [2]陕西省信息通信网络及安全重点实验室,陕西西安710121

出　　处：《西安邮电大学学报》2025年第1期19-26,共8页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金项目(61901366);陕西省自然科学基础研究计划项目(2020JQ-851);陕西省普通高校青年杰出人才支持计划项目。

摘　　要：针对无人机(Unmanned Aerial Vehicle,UAV)在目标检测过程中面临的通信资源浪费和效率低的问题,提出一种基于知识图谱(Knowledge Graph,KG)的UAV目标检测认知语义通信系统。利用KG知识和全局语义推理能力构建系统推理和传播算法,纠正传输过程中符号层面出现的错误,并采用注意力机制算法解决KG种类与实际图像中种类不平衡的现象。将所提系统与其他相关的通信系统在不同信噪比和不同信道环境下的检测精度和图像重建性能等进行对比,验证结果表明,所提系统在检测精度和图像重建方面均优于基准通信系统,在5 dB信噪比的加性高斯白噪声信道下,所提系统目标检测平均精度达到53.17%,相比其他系统提升了约15.6个百分点;图像重建结构相似性达到80.17%,提升了约10个百分点。To mitigate communication resource wastage and inefficiency in UAV-based target detection,a knowledge graph(KG)-based cognitive semantic communication system is proposed.KG knowledge and global semantic inference are adopted to construct the system inference and propagation algorithm,so as to rectify the symbol-level transmission errors.An attention mechanism is used to solve the class imbalance between the KG types and the actual image types.The proposed system is compared with other relevant communication systems under different signal-to-noise(SNR)ratios and channel conditions in terms of detection accuracy and image reconstruction performance.Verification results show that the proposed system is superior to the reference communication system in terms of detection accuracy and image reconstruction.Under the 5dB SNR additive white Gaussian noise channel,the proposed system achieves an average target detection accuracy of 53.17%,which is about 15.6 percentage higher than the other systems.The structural similarity index measure(SSIM)for image reconstruction reaches 80.17%,with an increase of about 10%.

关 键 词：无人机 目标检测 认知语义通信 知识图谱 注意力机制 语义纠错 

分 类 号：TN92[电子电信—通信与信息系统]