Unmanned Aerial Vehicle Intrusion Detection Method Based on Multi-scale Hierarchical Pyramid Network
In recent years,the rapid development of drones has brought convenience to many fields,but the intrusion of these drones can bring huge challenges for the security of airports.Due to the small target size,complex background,and fast flight speed of drones,existing mainstream methods are difficult to ac-curately identify drones,leading to false detection and missed detection.This paper proposes a novel multi-scale hierarchical pyramid network for unmanned aerial vehicle intrusion detection.The proposed meth-od utilizes feature fusion modules to assign semantic information of images at different levels and scales to the feature pyramid.Through grid deletion and 4-Mosaic data augmentation technology,the small sample dataset is expanded to improve the generalization performance.The experiment shows that the proposed method is improved by 5.5％better compared to the existing object detection methods.
万方数据
维普
