Improving real-time drone detection for counter-drone systems

Abstract

The number of unmanned aerial vehicles (UAVs, also known as drones) has increased dramatically in the airspace worldwide for tasks such as surveillance, reconnaissance, shipping and delivery. However, a small number of them, acting maliciously, can raise many security risks. Recent Artificial Intelligence (AI) capabilities for object detection can be very useful for the identification and classification of drones flying in the airspace and, in particular, are a good solution against malicious drones. A number of counter-drone solutions are being developed, but the cost of drone detection ground systems can also be very high, depending on the number of sensors deployed and powerful fusion algorithms. We propose a low-cost counter-drone solution composed uniquely by a guard-drone that should be able to detect, locate and eliminate any malicious drone. In this paper, a state-of-the-art object detection algorithm is used to train the system to detect drones. Three existing object detection models are improved by transfer learning and tested for real-time drone detection. Training is done with a new dataset of drone images, constructed automatically from a very realistic flight simulator. While flying, the guard-drone captures random images of the area, while at the same time, a malicious drone is flying too. The drone images are auto-labelled using the location and attitude information available in the simulator for both drones. The world coordinates for the malicious drone position must then be projected into image pixel coordinates. The training and test results show a minimum accuracy improvement of 22% with respect to state-of-the-art object detection models, representing promising results that enable a step towards the construction of a fully autonomous counter-drone system.

Geographic information spatial feature extraction based on drone aerial surveying and mapping technology

“…After completing the aerial photography, it is necessary to process and analyze the collected image data to obtain the required geographic information data [5] . The collection of geographic information spatial images is easily affected by external environments, weather, buildings, or drones themselves.…”

Countering a Drone in a 3D Space: Analyzing Deep Reinforcement Learning Methods

“…In this survey, the different drone detection and defense systems based on different types of methods which were proposed in the literature are presented. In the real world, the target position can be detected in many different sensing technologies such as radar [4,5], acoustical [6,7], radio frequency (RF) [8], optical [9], lidar [10], or a deep learning-based solution [11,12].…”

“…Critical part of the counter-drone systems to eliminate the target such as the time to interact with the target and the actions which the drone can use are considered carefully. A detailed explanation of the drone detection model used in this paper is provided in our previous research [11] which proposes a drone detection model trained by using different kinds of images of drones to obtain a more robust drone detector. The last and relevant contribution is the explainability of deep reinforcement learning.…”

“…EfficientNet-B0 is adapted for small-size objects. A detailed explanation of our drone detection model is provided in our previous research [11].…”

Cooperative localisation of UAV swarm based on adaptive SA-PSO algorithm

“…Recently, the localisation of UAV swarms in the civil and military domains has become an important area of research [1,2,3,4,5]. Generally, depending on the combination of information from the inertial measurement unit (IMU) and global navigation satellite system (GNSS), each UAV can offer position information for tracking or guidance [6,7,8].…”