LTC-YOLO: An Efficient and Lightweight Tower Crane Detection Algorithm for Complex Airport Clearance Scenarios

Airport clearance operations are vital for maintaining flight safety. However, current airport clearance operations primarily depend on manual processes, which are not only cumbersome but also challenging to monitor continuously. This paper presents LTC-YOLO, a lightweight algorithm for detecting tower cranes in complex airport clearance scenarios. Firstly, we introduce the C3GhostNetV2 module, which reduces parameters while enhancing feature extraction capabilities, thereby improving the model’s detection performance in complex environments. Secondly, we implement the LAMP pruning method, significantly reducing parameters and boosting real-time performance. Finally, the model’s detection accuracy is further enhanced through knowledge distillation using the CWD algorithm. Experiments on the self-constructed TC-Dataset demonstrate that, compared to YOLOv8s, LTC-YOLO maintains accuracy while reducing parameters by $\mathbf{79.8} \%$, lowering complexity by $\mathbf{38.5 \%}$, and increasing detection speed by $57 \%$. The model fills the gap of tower crane detection in airport clear zones, meets the expected requirements, and effectively assists the monitoring of airport clear zones.