<p>Accurately obtaining airport objects is crucial to ensuring airport safety and improving airport efficiency. The previous research has found that airport scene images have so many small objects. The presence of hazy reduces visibility and affects the ability to obtain information about objects in airport scene images. To address the challenges, this article proposes a small object detection based on improved YOLOv8 for airport scene images in hazy weather (ADH-YOLO). Firstly, this article constructs HASS1, HASS2, and HRSOD hazy datasets based on the HAZERD method. We find that YOLOv8 which is anchor-free detection shows significant performance improvement after using anchor boxes in hazy datasets. Then, to adapt to hazy environments, we design a decoupled detection head with the attention module (DDAH) and add a coordinate attention (CA) module to the backbone network. A small object detection layer is added to the original structure to further improve the performance of small object detection. Finally, this article uses partial convolution (PConv) to balance detection performance and computational resource consumption to reconstruct the backbone and neck C2F modules. We propose two lite versions of ADH-YOLO (LADH-YOLOn and LADH-YOLOa) compared with the original YOLOv8x, in which Params decreased by 22.1% and 49.6%, with higher mAP values. The proposed method improves mAP by 10.5% and 31.2% on the HASS1 and HASS2. Generalization experiments show that the proposed method achieves 96.1% mAP optimal detection results compared to classical detection methods on the HRSOD. The source code is available at <a href="https://github.com/rookie257/HAD-YOLO">https://github.com/rookie257/HAD-YOLO</a>.</p>