Aircraft Detection Based on Improved YoloV11

With the increasing diversity and complexity of aircraft types, accurate and efficient target detection technology has become an important demand in the field of target recognition. This paper proposes a multi-class aircraft detection method based on improved YOLOv11. By introducing two key modules, ADown module and Multi-Scale Dilated Attention(MSDA), the robustness and computational efficiency of the model are significantly improved while ensuring detection accuracy. Firstly, in order to solve the problem that YOLOv11 may lose important features during the downsampling process, this paper introduces the ADown module. This module uses a more detailed feature processing method to optimize the model's downsampling process by dynamically adjusting and reconstructing different feature layers, thereby enhancing the model's ability to capture details in complex scenes. Secondly, combining the dilated convolution and Transformer structure, this paper introduces the Multi-Scale Dilated Attention(MSDA) module. This module effectively improves the detection accuracy of different types of aircraft in complex backgrounds by adaptively adjusting the attention mechanism at different scales, especially in the processing of shallow features and local details. In this paper, a multi-class dataset containing 77 different aircraft is used and compared with the traditional YOLOv11 model and other state-of-the-art target detection algorithms. The experimental results show that the improved YOLOv11 model outperforms the baseline model in multiple evaluation metrics and excels in precision and recall. In summary, the improved method proposed in this paper shows strong advantages in various aircraft detection tasks, It can effectively cope with the detection challenges of a wide variety of targets with similar features and complex backgrounds, and has good application prospects.