Unmanned aerial vehicle (UAV) high-altitude small target identification holds application significance in urban construction, military reconnaissance, and maritime surveillance. To address challenges such as tiny targets, complex backgrounds, and occlusions in aerial images, this study proposes the APW-YOLOv8 algorithm. The algorithm introduces an ASCPA mechanism, which utilizes SPE to extract multi-scale spatial information and employs SCFM to achieve deep fusion of shallow and deep features, significantly enhancing the model’s perception of multi-scale information. Additionally, APW-YOLOv8 incorporates a P2 small target detection layer in the Neck network, which concatenates shallow and deep feature maps to enhance the extraction ability of small target features. Furthermore, the Inner-CIoU loss function is introduced to optimize loss computation by combining the scale factor, thereby improving multi-scale feature extraction performance. On the VisDrone2019 dataset, mAP@0.5 and mAP@0.5:0.95 of APW-YOLOv8 reach 31.9% and 18.8% respectively, an improvement of 3.7% and 2.3% respectively over YOLOv8n. On the ultra-long-range remote sensing RSOD dataset, mAP@0.5 and mAP@0.5:0.9 of APW-YOLOv8 increase by 3.7% and 2% respectively compared to YOLOv8n. Experimental results demonstrate that APW-YOLOv8 effectively solves missed and false detections of small targets, providing an efficient and practical solution for complex airspace detection, with broad application prospects.