在计算机视觉应用中，基于孪生网络的跟踪算法相比于传统的目标跟踪算法在速度和精度上都有所提升，但是其受到遮挡、形变等干扰因素影响较大。基于此，本文对现有基于孪生网络的目标跟踪方法和技术所作的改进进行了总结分析，主要包括在孪生网络中引入全卷积孪生神经网络方法、引入回归方法和在线更新方法，对基于3种方法的目标跟踪算法的改进进行了综述，并详细介绍了近年来孪生网络在目标跟踪应用中的国内外研究进展和发展现状。同时，采用 VOT2017和LaSOT数据集进行了实验对比，比较了多种基于孪生神经网络跟踪算法的性能。最后，对基于孪生网络的目标跟踪方法的发展趋势进行了展望。

In the computer vision applications, the tracking algorithms based on the Siamese networks have improved in speed and accuracy compared with the traditional target tracking algorithms, but they are greatly affected by interference factors such as occlusion and deformation. Based on these, the existing target tracking methods and technologies based on Siamese networks are summarized and analyzed, they mainly include the introduction of fully convolutional Siamese neural network method, regression method and online update method in Siamese networks, and the improvements of the target tracking algorithms based on three methods are reviewed. The research progress and development status of Siamese networks in target tracking applications in recent years are introduced in detail. Then, the VOT2017 and LaSOT datasets are used for experimental comparison, and the performances of various tracking algorithms based on Siamese neural networks are compared. At the end, the development trend of the target tracking methods based on Siamese neural networks is prospected.