移动通信2025，Vol.49Issue(7)：77-84,8.DOI:10.3969/j.issn.1006-1010.20250525-0002
面向多源视觉的语义传输方法研究
Research on Semantic Transmission Methods for Multi-Source Vision
摘要
Abstract
With the rapid development of machine perception,edge computing,and artificial intelligence,scenarios such as intelligent transportation and low-altitude unmanned systems pose increasing demands on the efficient transmission and processing of multi-sensor visual data.Semantic communication,as a new paradigm that integrates communication and intelligence,offers a novel approach to handling multi-source visual information.This paper first outlines the key challenges in multi-source vision processing,including modality misalignment,inefficient fusion,and limited channel adaptability.Then,the current research progress and representative methods for semantic transmission of multi-source vision are reviewed.Focusing on two representative application scenarios—intelligent vehicular networks and UAV-based platforms—a semantic transmission architecture is proposed for multi-source visual data.Specifically,a semantic transmission scheme for RGB-infrared image fusion is designed to address the demands of efficient transmission and adaptive processing.By integrating RGB—infrared semantic feature fusion strategies,the proposed approach enables efficient extraction and transmission of semantic content across modalities.Experimental results on the VTUAV dataset demonstrate that the proposed method maintains high reconstruction quality and task-relevant performance under low signal-to-noise ratio conditions,significantly outperforming traditional syntax-based transmission in both image reconstruction and tracking accuracy.关键词
多源视觉/语义通信/图像传输Key words
multi-source vision/semantic communication/image transmission分类
信息技术与安全科学引用本文复制引用
王佳琪,郑思源,刘宜明..面向多源视觉的语义传输方法研究[J].移动通信,2025,49(7):77-84,8.基金项目
北京市自然科学基金资助项目"复杂工业网络下面向语义通信的端边协同训推优化方法研究"(L251036) （L251036）
国家自然科学基金"语义信息的表征与传输理论"(62293481) （62293481）
中央高校基本科研业务费专项资金资助项目"算网融合的可信边缘协同计算关键技术研究"(2023RC95) （2023RC95）
