Efficient traffic management is vital for reducing congestion, improving road safety, and enabling smart city planning. This study compares three object detection models, YOLOv5, YOLOv8, and YOLO11, evaluating their performance in real world traffic conditions, including daytime, nighttime, rain and fog. The focus was on accurately detecting partially occluded and densely packed vehicles, common challenges in urban traffic scenes. Performance metrics such as mean Average Precision (mAP), precision, recall, and accuracy were used to assess each model. Results show that YOLO11 outperforms YOLOv5 and YOLOv8 consistently, demonstrating higher accuracy and robustness across different environmental conditions. YOLOv8 improved on YOLOv5 but still lagged behind YOLO11. This analysis highlights YOLO11 as the most reliable choice for real-time vehicle detection applications, making it ideal for smart traffic monitoring systems that require dependable performance in complex and variable traffic scenarios.