Due to the imaging specificity of infrared images, compared with the RGB images, there are drawbacks such as blurring, low contrast, unclear texture details and large background noise. So the tracking is more difficult and challenging. There are two main methods for infrared target tracking. One is to use Handcrafted features, but it is time-consuming and costly. The other is infrared target tracking algorithm which is the current mainstream algorithm and is to apply the tracking model used for RGB tracking to infrared target tracking. Compared with the RGB images, the image features extracted from infrared images often have low quality and poor tracking performance. Therefore, to address these problems, based on the idea of Siamese network, an efficient infrared target tracking model SiamMCA is designed in this paper. SiamMCA is mainly composed of a feature extraction network module, a mixed channel feature enhancement module, a multi-scale feature fusion module and a regional regression network module (RPN) for object classification and bounding box regression. Firstly, the feature extraction network module based on the deep residual network Resnet50 extracts richer and more comprehensive feature information by introducing multi-head self-attention. Then the mixed channel feature enhancement module assigns different weights to the features of different channels, which makes the model more focus on the regions of interest. Finally, response maps with multi-scale features are obtained by a multi-scale feature fusion module and are fed into the regional regression network module to classify the target and predict and regress the bounding box. The experimental results show that SiamMCA achieves nice performance on infrared data benchmarks LSOTB-TIR and PTB-TIR, and can effectively cope with various infrared target tracking tasks.