RGBT多模态视觉跟踪方法综述

Survey of RGBT Multimodal Visual Tracking Methods

摘要:

RGBT视觉跟踪是指融合可见光和热红外多模态图像信息进行视觉跟踪的新兴热点研究课题，合理融合可见光和热红外图像的互补信息可以提高跟踪器的性能和鲁棒性。人工智能技术的发展推动了RGBT多模态视觉跟踪的发展，深度学习技术逐渐代替传统目标跟踪方法，在精确度与速度方面更具有优势。对近年来RGBT多模态视觉跟踪进行了全面综述，整理了RGBT多模态视觉跟踪的发展历程，归纳和讨论了相关算法，具体包括基于相关滤波的方法和基于深度学习的方法；回顾了RGBT多模态视觉跟踪数据集的发展历史，介绍了算法性能评估指标，分析了不同方法在评估数据集上的性能，展望了RGBT多模态视觉跟踪的未来研究趋势。旨在为相关研究者提供全面的概览和参考，以促进RGBT多模态视觉跟踪领域的研究和发展。

Abstract:

RGBT visual tracking refers to the emerging hot research topic of fusing visible and thermal infrared multimodal image information for visual tracking, and the reasonable fusion of complementary information of visible and thermal infrared images can improve the performance and robustness of trackers. Artificial intelligence technology has promoted the development of RGBT multimodal visual tracking, and deep learning technology gradually replaces the traditional target tracking method, which has more advantages in terms of accuracy and speed. This paper is presented to organize the development history of RGBT multimodal visual tracking, summarize and discuss related algorithms, specifically including correlation filtering-based methods and deep learning-based methods, review the development history of evaluation datasets, introduce algorithm performance evaluation metrics, analyze the performance of different algorithms on evaluation datasets, and look forward to the future research trends of RGBT multimodal visual tracking methods with deep learning are presented. This paper aims to provide a comprehensive overview and reference for related researchers to promote research and development in RGBT multimodal vision tracking.