Abstract
Precise intrusion unmanned aerial vehicle (UAV) detection over long distances is of crucial importance for guaranteeing the low-air security. Although many deep learning-based vision detectors have been developed, they still rely on a large amount of hand-crafted fixed feature priors. Thus, the existing static dense-based detectors suffer from the severe mismatch and imbalance between the small size and high mobility of UAVs. To solve the problem, a novel vision-radar fusion-based dynamic sparse network (Vira-DSNet) is proposed for more balanced and precise UAV detection. The Vira-DSNet exploits our designed dynamic sparse candidate generator and radar-guided semantic feature transform module to generate a small set of customized high-quality object candidates and UAV semantic features based on the radar data. Furthermore, based on Hungarian bisection matching, the proposed Vira-DSNet eliminates the post-processing and is completely end-to-end differentiable. Moreover, the Vira-DSNet is deployed in our developed actual vision-radar fusion-based early intrusion UAV detection system to evaluate the performance in practical applications. Experimental results demonstrate that our Vira-DSNet achieves an average precision AP 50 of 88.2%. It is also shown that the average recall AR 1 of Vira-DSNet is higher than the state-of-the-art scheme by 10.1 %.