Research Article
Quantum-Enhanced Vision Transformer for Ultra-Fast and Precise Forest Fire Detection in UAV Surveillance
@INPROCEEDINGS{10.4108/eai.28-4-2025.2357859, author={Challa Venkata Sai Priya and Banda Pooja and Golla Nandini and Erapogu Preethi and Gangarapu NagaLakshmi}, title={Quantum-Enhanced Vision Transformer for Ultra-Fast and Precise Forest Fire Detection in UAV Surveillance}, proceedings={Proceedings of the 4th International Conference on Information Technology, Civil Innovation, Science, and Management, ICITSM 2025, 28-29 April 2025, Tiruchengode, Tamil Nadu, India, Part I}, publisher={EAI}, proceedings_a={ICITSM PART I}, year={2025}, month={10}, keywords={quantum machine learning vision transformers forest fire detection uav surveillance real-time inference dynamic attention mechanism}, doi={10.4108/eai.28-4-2025.2357859} }- Challa Venkata Sai Priya
Banda Pooja
Golla Nandini
Erapogu Preethi
Gangarapu NagaLakshmi
Year: 2025
Quantum-Enhanced Vision Transformer for Ultra-Fast and Precise Forest Fire Detection in UAV Surveillance
ICITSM PART I
EAI
DOI: 10.4108/eai.28-4-2025.2357859
Abstract
Early detection of forest fire is of vital importance to reduce environment and economic damages caused by forest fire. In this paper we propose Quantum-Dynamic Attention Vision Transformer (QDA-ViT): a new framework for real-time and resource efficient wildfire detection using unmanned aerial vehicles (UAVs). The model proposed integrates quantum probabilistic routing into the transformer’s attention mechanism so the attention head selection can be dynamically done via entropy driven spatial volatility. This increases focus of the regions where scattering happens and decrease the computational overhead. Moreover, a patch level compression module greatly lowers the data transmission load and thus makes the system applicable for onboard UAV processing. QDA-ViT shows better accuracy (96.3%), F1-score (0.902) and runtime (18.2 ms), than conventional CNN based and standard Vision Transformer on test results, which can prove its usefulness for real time online aerial surveillance for wildfire cases.
