Role of AI in Anti-drone Systems: A Review

Unmanned Aerial Vehicles (UAVs), commonly known as drones, are ubiquitous due to their diverse applications, from aerial photography to agriculture and disaster management. However, their widespread usage has also introduced new security challenges, especially to critical installations like nuclear facilities. The potential threats posed by drones, such as unauthorized aerial surveillance, hazardous payload dropping or even collision, raise significant concerns. To safeguard critical facilities against these threats, there is an immediate requirement for effective drone detection and tracking techniques. This paper reviews various modalities like RADAR, Radio Frequency (RF), Acoustic or Electro-Optical/Infrared (EO/IR) sensors that have been effective in detecting and localise drones. However, this paper focuses more on the existing literature on techniques based on EO/IR approaches. The advancements of Artificial Intelligence (AI) and Deep learning (DL) in various tasks of computer vision can be leveraged for EO/IR images. DL can be utilized to train models to differentiate drones from other objects, minimizing false alarms. A deep learning model based on CNNs is trained for differentiating birds and drones in aerial images, and the preliminary results are presented. This paper also highlights the open research issues in the domain of drone detection and localisation.