Drone image object detection is a crucial foundational technology across multiple research fields. However, due to challenges such as small object sizes, dense distributions, and complex backgrounds commonly encountered during drone-based object recognition, the detection accuracy is often suboptimal. To address these challenges, we optimize YOLOv8 and proposed a new object detection model specifically designed for drone aerial imagery, named YOLO-LiteMax. Firstly, we replace the Bottleneck structure in the Cross Stage Partial (CSP) module with a FasterNet Block structure, which effectively reduces parameter while maintaining the original level of accuracy. Secondly, we introduce a new feature pyramid structure called Small Target Scale Sequence Fusion (STSSF), enhancing multi-scale feature fusion to better capture small objects. Lastly, we replace the original detection head with the Shared Convolution Precision Detection (SCPD) head, which uses shared convolutions and group normalization to improve the efficiency and consistency of feature extraction. Experiments on the VisDrone2019 dataset show that YOLO-LiteMax significantly outperforms other baseline methods in detection accuracy. Compared to YOLOv8, YOLO-LiteMax achieved a remarkable 5.9% improvement in the mAP50 metric. These results demonstrate that YOLO-LiteMax significantly improves small object detection in drone imagery, making it more effective in complex environments.