pen icon Colloque
quote

Identification of Forest Fire Smoke from AVHRR Imagery by Means of Neural Networks

Alexandre Khananian
R. Fraser
Z. Li
AK

Membre a labase

Alexandre Khananian

Résumé du colloque

Satellite remote sensing offers a practical means of monitoring forest fires and smoke from biomass burning. Monitoring boreal forest fires is important to forest protection and the management of forest resources. It is also crucial for understanding global carbon budget in light of the greenhouse effect. Boreal forest fires inject a large amount of greenhouse gases and aerosol particles into the atmosphere. To assess the impacts of biomass burning, techniques must be developed for routinely measuring the spatial and temporal distribution of fires and smoke on regional and global scales. In the present study, an extension of a satellite-based fire detection algorithm (Li et al., 1997) was developed to identify smoke across the Canadian boreal forest. The automated algorithm uses a neural network approach for identifying dense smoke plumes from NOAA-AVHRR (Advanced Very High Resolution Radiometer) imagery. The technique uses a feed-forward neural network based on multilayer perceptron’s topology (MLP) with back-propagation learning. Two networks have been developed for different applications. The first (NN-1) was developed to provide a rapid procedure for screening smoke plumes from AVHRR imagery. The second (NN-2) uses multispectral information to classify smoke, cloud, land cover, and water components of each AVHRR scene. The method has been applied to a large area in central western Canada where extensive burning occurred in the past fire season, and will be extended for application across Canada. The method will form a key component in a comprehensive, satellite-based forest fire monitoring system currently being developed at the Canada Centre for Remote Sensing. A comparison with other smoke detection methods (e.g. simple channel threshold) is also performed.

Contexte

Section :
La télédétection optique et radar et la géomatique pour la gestion des problèmes environnementaux
news icon Thème du colloque :
La télédétection optique et radar et la géomatique pour la gestion des problèmes environnementaux
manager icon Responsables :
Abdou Bannari
host icon Hôte : Université d’Ottawa

Découvrez d'autres communications scientifiques

news icon

Titre du colloque :

La télédétection optique et radar et la géomatique pour la gestion des problèmes environnementaux
Étude environnementale du secteur forestier Mintom-Lomie dans le Sud-Est du Cameroun (Afrique Centr…
Paul-Desire Ndjigui
Contribution à l'élaboration d'un SIG intégrant la télédétection pour la gestion et le suivi de la …
Mohamed Benmoussa
Intégration des données de télédétection multispectrale dans un SIG pour l'étude de la dynamique de…
Mohamed Cherkaoui
Voir tous les contenus de ce colloque

Autres communications du même congressiste :

news icon

Thème du colloque :

La télédétection optique et radar et la géomatique pour la gestion des problèmes environnementaux
Identification of Forest Fire Smoke from AVHRR Imagery by Means of Neural Networks
Alexandre Khananian
Identification of Forest Fire Smoke from AVHRR Imagery by Means of Neural Networks
Alexandre Khananian