Benchmarking of wildland fire color segmentation algorithms

T Toulouse (1) , Lucile Rossi (2) , M Akhloufi (3) , T Celik (4) , Xavier Maldague (4)

1 LVSN - Laboratoire de Vision et Systèmes Numériques de l'Université Laval
2 Feu
3 UTFSM - Electronics Department [Valparaiso]
4 WITS - University of the Witwatersrand [Johannesburg]

Recently, computer vision-based methods have started to replace conventional sensor based fire detection technologies. In general, visible band image sequences are used to automatically detect suspicious fire events in indoor or outdoor environments. There are several methods which aim to achieve automatic fire detection on visible band images, however, it is difficult to identify which method is the best performing as there is no fire image dataset which can be used to test the different methods. This paper presents a benchmarking of state of the art wildland fire color segmentation algorithms using a new 1 fire dataset introduced for the first time in this paper. The dataset contains images of wildland fire in different contexts (fuel, background, luminosity, smoke). All images of the dataset are characterized according to the principal color of the fire, the luminosity and the presence of smoke in the fire area. With this characterization, it has been possible to determine on which kind of images each algorithm is efficient. Also a new probabilistic fire segmentation algorithm is introduced and compared to the other techniques. Bench-marking is performed in order to assess performances of twelve algorithms that can be used for the segmentation of wildland fire images.

Mots clés: Image dataset; Benchmarking; Wildland fire; Color; Segmentation

Domaines: Traitement du signal et de l'image [eess.SP]

T Toulouse, Lucile Rossi, M Akhloufi, T Celik, Xavier Maldague. Benchmarking of wildland fire color segmentation algorithms. IET Image Processing, 2015, 9 (12), pp.1064-1072. ⟨10.1049/iet-ipr.2014.0935⟩. ⟨hal-01182107⟩