自动化学报2025，Vol.51Issue(12)：2664-2678,15.DOI:10.16383/j.aas.c250118
基于语义概念关联的参考多目标跟踪方法
Semantic Conceptual Association-Based Method for Referring Multi-Object Tracking
摘要
Abstract
Referring multi-object tracking(RMOT)is a task that jointly leverages language and visual modalities for object localization and tracking,aiming to accurately identify and continuously track specific objects in video frames according to natural language prompts.Although existing RMOT methods have achieved notable progress,their modeling of the conceptual granularity of language expressions remains limited,leading to insufficient semant-ic parsing when handling complex descriptions.To address this issue,we propose a semantic concept association-based RMOT framework,termed SCATrack.The framework introduces two key modules,namely the sharing se-mantic concept(SSC)module and the semantic concept auxiliary generation(SCG)module,to enhance the model's capability of deeply understanding language expressions,thereby improving both the continuity and robustness of tracking.Specifically,the SSC module performs semantic concept partitioning over the language expressions,en-abling the model to effectively distinguish different expressions conveying the same semantics and similar expres-sions across distinct semantics,which strengthens object discrimination under multi-granularity input conditions.The SCG module adopts a feature masking and generation mechanism to guide the model in learning representa-tions of multi-granularity language concepts,thereby improving its robustness and discriminative ability in com-plex language scenarios.Experimental results on two widely used benchmark datasets demonstrate that the pro-posed SCATrack significantly improves tracking performance in RMOT tasks,validating its effectiveness and su-periority.关键词
参考多目标跟踪/多模态融合/对比学习/目标跟踪/提示场景理解Key words
Referring multi-object tracking/multi-modal fusion/contrastive learning/object tracking/prompt scene understanding引用本文复制引用
LIN Jia-Cheng,CHEN Jia-Jun,LI Zhi-Yong,WANG Yao-Nan..基于语义概念关联的参考多目标跟踪方法[J].自动化学报,2025,51(12):2664-2678,15.基金项目
国家自然科学基金(U23A20341),湖南省重大科技攻关计划(2025QK1005)资助 Supported by National Natural Science Foundation of China(U23A20341)and Major Scientific and Technological Research Plan of Hunan Province(2025QK1005) （U23A20341）
