This paper introduces a smart city management system using a real-time target detection algorithm on Unmanned Aerial Vehicles to address urban road occupation issues. Equipped with a high-resolution camera, the UAV collects data that undergo preprocessing for model training. During law enforcement, the drone camera's video is streamed in real time to relevant equipment via a video acquisition device, which processes the video using the YOLOv8 target detection algorithm and displays the results. YOLOv8, a lightweight yet high-performance model in the YOLO series, employs a unique dual-path prediction and closely connected convolutional network for target detection. It features high efficiency and can handle targets of varying sizes through cascade and pyramid concepts. The model, trained on preprocessed data, achieves outstanding results with a 95.1% debris recognition accuracy, 97.2% recall rate, and 99.1% mAP50. This model demonstrates excellent classification capabilities.