This paper presents a comprehensive evaluation of YOLO object detection models for detecting aircraft, ships, and motor vehicles using UAV-based aerial imagery within the framework of smart cities. Efficient and accurate object detection is critical for real-time surveillance, traffic monitoring, and intelligent transportation systems (ITS). The study compares key performance metrics of each model under challenging conditions such as varying light, occlusions, and diverse object orientations. Among the tested models, YOLOv8 demonstrated superior performance, achieving the highest precision across all object categories, particularly excelling in ship detection due to the clear distinction between objects and their backgrounds. Furthermore, it achieved the fastest inference times, processing about 101 images per second, making it highly suitable for real-time applications. Despite these strengths, YOLOv8 faced challenges with certain complex scenes, such as overlapping objects, shadow interference, and small object detection. Nevertheless, its overall efficiency and accuracy establish it as the most promising model for deployment in remote sensing-based monitoring systems in smart cities.