Drone-type-Set: Drone types detection benchmark for drone detection and tracking

Kholoud AlDosari,AIbtisam Osman,Omar Elharrouss,Somaya AlMaadeed,Mohamed Zied Chaari

2024-05-17

Abstract:The Unmanned Aerial Vehicles (UAVs) market has been significantly growing and Considering the availability of drones at low-cost prices the possibility of misusing them, for illegal purposes such as drug trafficking, spying, and terrorist attacks posing high risks to national security, is rising. Therefore, detecting and tracking unauthorized drones to prevent future attacks that threaten lives, facilities, and security, become a necessity. Drone detection can be performed using different sensors, while image-based detection is one of them due to the development of artificial intelligence techniques. However, knowing unauthorized drone types is one of the challenges due to the lack of drone types datasets. For that, in this paper, we provide a dataset of various drones as well as a comparison of recognized object detection models on the proposed dataset including YOLO algorithms with their different versions, like, v3, v4, and v5 along with the Detectronv2. The experimental results of different models are provided along with a description of each method. The collected dataset can be found in

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

This paper focuses on the detection and tracking of unmanned aerial vehicles (UAVs), especially those used for illegal activities such as drug trafficking, espionage, and terrorist attacks. Due to the widespread use and low cost of UAVs, preventing these potential threats has become crucial. The paper presents a challenge of identifying unauthorized UAVs of different types, as there is currently a lack of datasets for such UAVs.

To address this issue, the authors created a large-scale dataset of UAV types, which includes images of different scales and perspectives. They evaluated various deep learning models, including YOLOv3, YOLOv4, YOLOv5, and Detectronv2. The experimental results include precision, recall, F1 score, and other metrics of the models. They also discussed the collection, partitioning, and configuration of the dataset, as well as the training process and outcomes.

The paper mentioned that existing methods for UAV detection often fail to identify the specific types of UAVs, and the size of the dataset is limited, which restricts the detection performance. Therefore, this new dataset provides detailed information on UAV types, which helps improve the accuracy and robustness of detection.

By comparing the performance of different models on multiple datasets, the study found that the YOLO series algorithms and Detectronv2 have good performance in UAV detection, but the specific optimal model may vary depending on the dataset. These research findings make significant contributions to improving UAV detection systems, especially the ability to identify different types of UAVs.