A MobileNetV2-CBAM-based Model for Forest Fire Classification Using UAV Imagery

Forest fires are one of the worldwide severe natural disasters causing human and environmental losses. Early and accurate fire detection is critical to minimize damages caused by forest fires and reduce human risks in firefighting efforts. Recently, Unmanned Aerial Vehicles (UAVs) mounted with camer...

Full description

Saved in:
Bibliographic Details
Published in2023 IEEE 4th International Conference on Pattern Recognition and Machine Learning (PRML) pp. 141 - 146
Main Authors Deng, Xinjie, Khan, Burhan, Lim, Chee Peng, Liao, Ming-Yu
Format Conference Proceeding
LanguageEnglish
Published IEEE 04.08.2023
Subjects
aerial imagery
attention mechanism
Autonomous aerial vehicles
Data models
forest fire classification
Forestry
Image coding
MobileNetV2
Power demand
Real-time systems
Training data
Online AccessGet full text

Cover

More Information
Summary:Forest fires are one of the worldwide severe natural disasters causing human and environmental losses. Early and accurate fire detection is critical to minimize damages caused by forest fires and reduce human risks in firefighting efforts. Recently, Unmanned Aerial Vehicles (UAVs) mounted with cameras and edge computing devices with Artificial Intelligence (AI)-based algorithms have been employed to tackle forest fire detection problems in real time. Considering that edge computing is limited in resources and power consumption, the developed AI algorithms need to be lightweight. This paper proposes lightweight deep learning (DL) algorithms based on MobileNetV2 and attention mechanism to detect forest fires using UAV imagery. Based on several experiments, the proposed model yields an accuracy rate of 86.14% and an F1-score of 0.8719, outperforming other related methods in forest fire detection and classification.
DOI:10.1109/PRML59573.2023.10348345