%0 Journal Article %A WANG Ling %A WANG Jiapei %A WANG Peng %A SUN Shuangzi %T Siamese Network Tracking Algorithms for Hierarchical Fusion of Attention Mechanism %D 2021 %R 10.3778/j.issn.1002-8331.2001-0330 %J Computer Engineering and Applications %P 169-174 %V 57 %N 8 %X

Based on the full-convolution Siamese network tracking algorithm SiamFC, this paper proposes a Siamese network target tracking algorithm fused attention mechanism. In the template branch, the neural network can learn the channel correlation and the spatial correlation of the template image through the attention mechanism fusion, thus increasing the foreground contribution, suppressing the background features, and improving the discrimination of network to positive samples features. Meanwhile, the VggNet-19 network is used to extract the shallow and deep features of the template image, the two features fuse adaptively. The experimental results on the datasets of OTB2015 and VOT2018 demonstrate that compared with SiamFC, the proposed algorithm can more effectively deal with the tracking problems, such as motion blur, target drift and background clutter, obtains higher accuracy and success rate.

%U http://cea.ceaj.org/EN/10.3778/j.issn.1002-8331.2001-0330