This internet browser is outdated and does not support all features of this site. Please switch or upgrade to a different browser to display this site properly.
Apply

Smoke Detection

Created a pseudo-hyperspectral labelled fire smoke training dataset, and derived a novel smoke detection algorithm tailored for light-weight applications requiring automated detection and minimal band selection.

HomeResearch at CurtinSSTCSmoke Detection

Research activities

This SmartSat CRC funded research project undertaken from March 2022 – October 2023 focused provide a solution for energy-efficient AI-based onboard processing of hyperspectral imagery for early fire smoke detection, as part of the Kanyini satellite mission pipeline.  The key activities related to the Curtin node staff (work formerly done while at UniSA) were generation of a comprehensive hyperspectral training dataset labelled with “Smoke”, “Cloud”, “Mixed”, and “Clear” and derivation of a novel  “HyperScout-2 Smoke Detection Algorithm” (HSSDA) for automated detection of smoke and delineation from other atmospheric aerosols.

Explore more

Publication

Check out our publication on adapting a lightweight convolutional neural network model to fit the computational constraints of the upcoming Kanyini mission with Hyperscout-2 camera.  We show that onboard AI for fire smoke detection with our algorithms can significantly improve downlink efficiency, energy consumption, and detection speed of Australian bushfires.

S. Lu et al., “Onboard AI for Fire Smoke Detection Using Hyperspectral Imagery: An Emulation for the Upcoming Kanyini Hyperscout-2 Mission,” in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 17, pp. 9629-9640, 2024, doi: 10.1109/JSTARS.2024.3394574.

Technical Report

Read about our research in our technical report.

SmartSat 2023, Energy efficient onboard AI for early fire-smoke detection, SmartSat Technical Report no. 11, SmartSat, Adelaide, Australia.  Authors: Stefan Peters, Sha Lu, Eriita Jones, Jixue Liu, Jiuyong Li, Jim O’Hehir (University of South Australia); Kai Qin, Yu Sun (Swinburne University); Norman Mueller, Simon Oliver (Geoscience Australia).

Article

Read more on this work with an article published by Space Connect, “New technique revolutionises bushfire detection from space”, 05 June 2024.