In the field of image fusion,existing methods based on Convolution Neural Network(CNN)or Transformer architectures have two limitations:first,effective aggregation between shallow texture features and deep semantic features is not possible；second,the weight ratio of infrared and visible light features cannot be adaptively changed.In this paper,an adaptive fusion method of infrared and visible images that introduce feature interaction is proposed.Firstly,a feature interaction module based on Transformer is constructed to aggregate cross scale feature information and enhance feature representation.Secondly,a fusion module is designed to adaptively adjust the feature weight ratio.The proposed fusion method is completed by a two stage training strategy.In the first stage,the encoder is trained using innovative feature interaction concepts to enhance feature representation and reconstruct feature images.In the second stage,the weight adaptive adjustment module based on the design trains the infrared and visible feature fusion task.The experimental results on publicly available datasets show that this method is superior to other typical methods in terms of subjective and objective evaluation compared with the existing methods.