Object Detection Algorithm of Aircrafts in Airport Scene Based on Improved YOLOv5

In order to solve the problems of false or missed object detection of aircrafts in complex airport scene, which is caused by obstruction and poor lighting, this paper proposes an improved YOLOv5 object detection algorithm in airport scene. Firstly, a Contextual Transformer Networks (CoT) is introduced in the feature extraction module. Secondly, a detection layer is added to the feature fusion module to improve the ability of perception of detailed features. Finally, decoupling is performed in the detection head. Different branches is used for classification and regression tasks to improve detection performance. The experiments use the Aircraft Dataset Airport Scene for model training and validation. The results show that our improved algorithm achieves 94.1% and 92.3% of the mAP@0.5 and mAP@0.5:0.95, which are 2.8% and 7.5% higher than the original YOLOv5 algorithm, respectively. At the same time, it ensured the detection speed well and met the real-time requirements.