10 - Safety and Security
A NOVEL METHOD FOR DETECTING UAVS USING PARALLEL NEURAL NETWORKS WITH RE-INFERENCE
H. Stepien, Saab Aeronautics, Sweden; M. Bilger, Saab Surveillance, Sweden; H. Forsberg¹, B. Lindgren, Saab Dynamics, Sweden; J. Hjorth¹; ¹Mälardalen University, Sweden
In this paper, we suggest a method for detecting unmanned aerial vehicles using a parallel architecture consisting of three diverse convolutional neural networks, each with different properties. The system utilizes re-inference to increase detection capability. The results confirm the method works and show major improvements in total accuracy as well as lower mispredictions.