This study presents a comprehensive evaluation of the Single Shot Multibox Detector (SSD) and YOLOv4 models for aircraft detection in Unmanned Aerial Vehicle (UAV)-captured images under varying environmental conditions. The models were compared based on their training and validation loss values, detection accuracy, and robustness in different lighting and altitude scenarios. The results indicate that YOLOv4 exhibits a more efficient learning process, with faster decline in loss values and better generalization performance, making it suitable for real-time applications. However, the SSD model outperformed YOLOv4 in terms of detection accuracy, particularly under medium-altitude, low-light, and challenging conditions such as varying color patterns, shadows, and aircraft geometries. Despite facing detection errors in certain high-altitude images, where complex objects like passenger boarding bellows systems resemble aircraft, the SSD model demonstrated superior resilience and detection speed. In both sufficient and insufficient light conditions, the SSD model consistently outperformed YOLOv4 in terms of precision, recall, and mean average precision (mAP), showcasing its robustness across different environmental challenges. This paper highlights the strengths and limitations of both models, providing valuable insights for choosing the optimal model for UAV-based aircraft detection tasks.