【摘要】 近年来，基于孪生网络的目标跟踪算法由于在跟踪精度和跟踪效率之间能够实现良好的平衡而备受关注。通过对基于孪生网络的目标跟踪算法的文献进行归纳，对现有孪生网络目标跟踪算法进行了全面总结，对孪生网络的2个分支结构进行了讨论。首先，介绍了基于孪生网络目标跟踪的基本架构，重点分析了孪生网络中主干网络的优化，以及主干网络的目标特征提取问题。其次，对目标跟踪过程中的分类和回归2个任务展开讨论，将其分为有锚框和无锚框2大类来进行分析研究，通过实验对比，分析了算法的优缺点及其目标跟踪性能。最后，提出未来的研究重点：1)探索背景信息训练，实现场景中背景信息传播，充分利用背景信息实现目标定位。2)目标跟踪过程中，目标特征信息的更加丰富化和目标跟踪框的自适应变化。3)从帧与帧之间全局信息传播，到目标局部信息传播的研究，为准确定位跟踪目标提供支撑。更多还原

【Abstract】 In recent years, the target tracking algorithm based on Siamese network has attracted much attention because it can achieve a good balance between tracking accuracy and tracking efficiency.Through the intensive study of the literature of target tracking algorithm based on Siamese network, the existing target tracking algorithm based on Siamese network was comprehensively summarized.Firstly, the basic framework of target tracking was introduced based on Siamese network, and the optimized backbone network in Siamese network and its target feature extraction were analyzed.Secondly, the classification and regression tasks in the process of target tracking were discussed, which were divided into two categories of anchor frame and anchor-free frame.The advantages and disadvantages of the algorithm as well as the target tracking performance were analyzed through experimental comparison.Finally, the focus of future research is proposed as following: 1) Explore the training of background information, realize the dissemination of background information in the scene, and make full use of background information to achieve target positioning.2) In the process of target tracking, the target feature information is enriched and the target tracking frame is changed adaptively.3) Research from the global information transmission between frames to the target local information transmission provides support for the accurate target positioning and tracking.更多还原