基于孪生网络的目标跟踪算法研究进展

Research Progress of Target Tracking Algorithm Based on Siamese Network

目标跟踪是计算机视觉领域中最为核心的基础研究问题之一,其能够协同高层视频应用分析和研究,具有重要的理论价值、广泛的实用价值和多学科交叉性,成为学术界、工业界以及国家战略的关注焦点。由于跟踪场景复杂度高、干扰强,目标表观变化多样性以及多模态信息融合等因素,使得跟踪器需要均衡鲁棒性、准确性以及实时性等性能衡量指标。目前,已有很多工作从不同视角解决目标跟踪领域中的挑战,但是在多维度性能指标的衡量下,仍然不能很好地克服复杂场景下的跟踪问题。本文通过基于孪生网络的目标跟踪算法,回顾领域发展现状,探讨存在的挑战,展望未来值得关注的研究方向,为该领域未来的研究工作提供借鉴和参考。

Target tracking is one of the core basic research issues in the field of computer vision.Its performance can cooperate with the analysis and research of high-level video applications,which has important theoretical value,extensive practical value and interdisciplinary.Therefore,it has become the focus of academia,industry and national strategy.Due to the high complexity and strong interference of the tracking scene,the diversity of target apparent changes and multi-modal information fusion,the tracker needs to balance the performance measurement indicators such as robustness,accuracy and real-time.At present,a lot of work has been done to solve the challenges in the field of target tracking from different perspectives,but under the measurement of multi-dimensional performance indicators,it still can not solve the tracking problem in complex scenes.Through the research of target tracking algorithm based on Siamese network,this paper reviews the current development status of the field,discusses the existing challenges and looks forward to the research direction worthy of attention in the future,so as to provide a reference for the future research work in this field.