可见光与红外融合目标跟踪技术研究进展综述
2022,30(10):7-7
摘要：目标跟踪是计算机视觉领域中的一个重要的问题，自从被提出以来逐渐发展出了不同类型的技术方法。由于可见光与红外光在目标跟踪方面存在互补性，两者融合的目标跟踪技术在性能和鲁棒性上比传统目标跟踪方法更具有优势。近年来，人工智能技术的发展推动了可见光与红外融合目标跟踪技术的快速进步。面向可见光与红外融合目标跟踪技术的发展历程，首先针对可见光与红外融合领域经典方法和近年来出现的技术进行梳理和总结，特别是对近两年的研究进展进行了归纳整理，具体包括基于Transformer、基于注意力机制、基于时间序列、自适应融合和基于多模态编解码器的可见光与红外融合目标跟踪方法；然后介绍了可见光与红外融合数据集及目标跟踪的评价指标；最后对未来的发展方向作了展望。
关键词：目标跟踪;可见光与红外融合(RGBT);深度学习;Transformer;注意力机制
A Review of the Research Progress on Visible Light and Infrared Fusion Target Tracking
Abstract：Object tracking is an important problem in the field of computer vision, and different types of technical methods have been gradually developed since it was proposed. Due to the complementarity of visible and infrared light in target tracking, the fusion of the two has more advantages than traditional methods in performance and robustness. In recent years, the development of artificial intelligence has promoted the rapid progress of visible and infrared light fusion target tracking. Facing the development history of visible and infrared light fusion target tracking technology, first of all, the classic methods in the field of visible and infrared light fusion that have appeared in recent years are sorted out, especially the research progress in the past two years is summarized, which are based on Transformer, attention, time series, self-adaptive fusion and multi-modal codec; then the visible and infrared light fusion datasets and target tracking evaluation index are introduced; last, the future development of this field is prospected.
Key words：Target tracking; RGB-infrared fusion tracking (RGBT); Deep learning; Transformer; Attention
收稿日期：2022-06-29
基金项目：中国航天科技集团公司钱学森青年创新基金资助项目
