To address the problems that the existing Siamese network tracking algorithm conducts similarity matching by merely employing the features of the last layer of the backbone network and it is lack of effective template update strategy, a Siamese network tracking algorithm is proposed based on multi-layer feature fusion and adaptive template update.Firstly, a novel zero padding unit is developed by combining deep over-parameterized convolution, and the deeper foreground features and semantic background are extracted.Secondly, a novel global-local feature fusion module is proposed for fully aggregating the global and local information of shallow layer features and capturing rich superficial features and transitional features of the middle layer.An adaptive template update mechanism is used to online update the template.Assessment is made on public benchmark dataset to verify the effectiveness of the algorithm and the experimental results show that, the proposed algorithm achieves the accuracy of 0.878 and 0.588 on the OTB2015 and VOT2018 datasets respectively, and the average overlap rate on the GOT10K dataset reaches 0.526, outperforming other algorithms.