The CNN-based tracker achieves excellent performance, but there are few researches on tracking TIRsingle target. TIR images have blurred outlines, lack color and texture information, so it is difficult to extract strong feature information and distinguish interference when using an RGB tracker to directly track TIR targets, which will lead to poor tracking robustness. We suggest a TIR tracker based on a Siamese tracker and a Transformer to address these issues. Specifically, we improved a feature extraction network with channel attention. Based on Transformer, we propose submodules named CA. We use Transformer as a matching module between the feature map of the template and the search region. The CA module realizes the mutual enhancement of the two path information, finds the feature dependence of the template and the search region, which provides rich context information. And we used the LSOTB-TIR evaluation dataset to verify its performance, which results that our tracker having advanced performance.