Texture-based classification for characterizing regions on remote sensing images

Abstract : Remote sensing classification methods mostly use only the physical properties of pixels or complex texture indexes but do not lead to recommendation for practical applications. Our objective was to design a texture-based method, called the Paysages A PRIori method (PAPRI), which works both at pixel and neighborhood level and which can handle different spatial scales of analysis. The aim was to stay close to the logic of a human expert and to deal with co-occurrences in a more efficient way than other methods. The PAPRI method is pixelwise and based on a comparison of statistical and spatial reference properties provided by the expert with local properties computed in varying size windows centered on the pixel. A specific distance is computed for different windows around the pixel and a local minimum leads to choosing the class in which the pixel is to be placed. The PAPRI method brings a significant improvement in classification quality for different kinds of images, including aerial, lidar, high-resolution satellite images as well as texture images from the Brodatz and Vistex databases. This work shows the importance of texture analysis in understanding remote sensing images and for future developments
Type de document :
Article dans une revue
Journal of applied remote sensing, Bellingham, WA : SPIE, 2017, 11 (03), pp.036028. 〈10.1117/1.JRS.11.036028〉
Liste complète des métadonnées

https://hal-sde.archives-ouvertes.fr/hal-01614631
Contributeur : Yannick Brohard <>
Soumis le : mercredi 11 octobre 2017 - 10:43:13
Dernière modification le : vendredi 19 octobre 2018 - 15:26:10

Identifiants

Citation

Frederic Borne, Gaëlle Viennois. Texture-based classification for characterizing regions on remote sensing images. Journal of applied remote sensing, Bellingham, WA : SPIE, 2017, 11 (03), pp.036028. 〈10.1117/1.JRS.11.036028〉. 〈hal-01614631〉

Partager

Métriques

Consultations de la notice

176