Webber, Luke (2024) Real-Time UAS Detection in High-Security Environments. [USQ Project]
![[img]](https://sear.unisq.edu.au/style/images/fileicons/text.png) |
Text (Project – redacted)
Webber_L_Dissertation_Redacted.pdf Download (3MB) |
Abstract
This thesis investigates the development and evaluation of a real-time drone detection system designed for high-security environments. The core objective was to ensure reliable detection of unmanned aerial systems (UAS), with a target detection accuracy of at least 50%. To achieve this, the YOLO11 object detection framework was implemented and evaluated against its variants, YOLO11m, YOLO11l, and YOLO11x, while considering trade-offs between accuracy and computational efficiency. Model training and hyperparameter optimisation were performed on a Google Cloud virtual machine to leverage its computational capabilities.
The optimised YOLO11 model delivered a mean average precision (mAP) of 0.469, a peak F1 score of 0.45, and a precision of 0.735 for the Drone class. These results reflect a considerable improvement over the baseline YOLO11 models. However, certain limitations were observed, particularly difficulties in detecting larger drones and classifying birds due to dataset constraints. Although the model effectively detected UAS across all five real-time test datasets, challenges such as inconsistent tracking in video-based tests suggest areas for refinement. Improvements could include dataset augmentation to address drone size diversity and the integration of advanced tracking algorithms.
The research concludes the optimised YOLO11 model offers a promising, efficient solution for UAS detection, particularly in high-security contexts like prison perimeter surveillance. Future efforts should prioritise building dataset diversity, stabilising tracking performance, and optimising the model for secure environments. This will ensure the model’s readiness for real-world applications, where precision and reliability in drone detection are vital in maintaining security.
|
Statistics for this ePrint Item |
| Item Type: | USQ Project |
|---|---|
| Item Status: | Live Archive |
| Faculty/School / Institute/Centre: | Current – Faculty of Health, Engineering and Sciences - School of Engineering (1 Jan 2022 -) |
| Qualification: | Bachelor of Engineering (Honours) (Electrical and Electronics) |
| Date Deposited: | 18 Mar 2026 22:50 |
| Last Modified: | 18 Mar 2026 22:50 |
| Uncontrolled Keywords: | UAV; UAS; RPAS; Drone; Object Detection; Machine Learning; Neural Network; Prison; Security; Optuna; YOLO |
| URI: | https://sear.unisq.edu.au/id/eprint/53174 |
Actions (login required)
 |
Archive Repository Staff Only |
