“A significant progress has been made in the research on aircraft target model recognition. This study proposes an aircraft model recognition method that integrates target segmentation and keypoint detection to address the limitations of current deep learning techniques in fine-grained recognition tasks. This method achieves high-precision recognition of aircraft models by combining multi task deep neural networks, conditional random fields, and template matching algorithms. The experimental results show that compared with traditional algorithms and fully end-to-end deep learning methods, this method has higher accuracy and practicality. This study first utilized multi task deep neural network transfer learning technology to achieve the recognition of aircraft target object position, mask, and keypoint information. Subsequently, by integrating the aircraft target mask refinement algorithm of the conditional random field and the attitude adjustment algorithm based on key points, the boundaries of the identified targets were refined and the aircraft attitude was adjusted. Finally, based on the constructed aircraft model template library, the refined and processed aircraft mask information was matched with the template library to achieve aircraft target model recognition. The results of this study not only provide new solutions for aircraft target model recognition, but also open up new directions for the application of deep learning technology in fine-grained recognition tasks. In the future, this method is expected to play an important role in fields such as aviation safety and remote sensing image analysis.”