An Upgraded-YOLO with Object Augmentation: Mini-UAV Detection Under Low-Visibility Conditions by Improving Deep Neural Networks

Over the last few years, the manufacturing technology of mini-unmanned aerial vehicles (mini-UAVs), also known as mini-drones, has been experiencing a significant evolution. Thus, the early warning optical drone detection, as an important part of intelligent surveillance, is becoming a global research hotspot. In this article, the authors provide a prospective study to prevent any potential hazards that mini-UAVs may cause, especially those that can carry payloads. Subsequently, we regarded the problem of detecting and locating mini-UAVs in different environments as the problem of detecting tiny and very small objects from an aerial perspective. However, the accuracy and speed of existing detection algorithms do not meet the requirements of real-time detection. For solving this problem, we developed a mini-UAV detection model called Upgraded-YOLO based on the state-of-the-art object detection method of YOLOv5. The proposed model is able to perform real-time tiny/small flying object detection. The main contributions of this research are as follows: firstly, an air image dataset of mini-UAVs was built using a Dahua multisensor camera. Secondly, a strategy of instance augmentation is proposed, in which we added small appearance of mini-drones to samples of the custom air image dataset. Thirdly, in addition to hyperparameter tuning and optimization operations, shallow layers are added to improve the model’s ability to detect mini-UAVs. A comparative study with several contemporary object detectors proved that the Upgraded-YOLO performed better. Therefore, the proposed mini-UAV detection technology can be deployed in a monitoring center in order to protect a strategic installation even in low-visibility conditions.