A Real-time Robust Approach for Tracking UAVs in Infrared Videos
Han Wu, Weiqiang Li, Wanqi Li, Guizhong Liu
Abstract
Object tracking has been studied for decades, but most of the existing works are focused on the RGB tracking. For an infrared video, the object is often textureless, especially for far-range drone planar targets. Furthermore, motion of camera and unexpected movement of the drones make tracking more difficult, causing existing object tracking algorithms lose the targets. In this paper a robust and realtime tracking algorithm is proposed for infrared drones, in which a feature attention module and an expansion strategy for searching the target are added to the fully convolutional classifier. Experiments on the Anti-UAV infrared dataset show its robustness to the different challenges of real infrared scenes with a high efficiency.