Robust classification of texture land forest inventory based on model of minimally sufficient features

• Main Authors: Ipatov Yury, Krevetsky Alexandr, Andrianov Yury, Sokolov Boris
• Format: Article
• Language: English
• Published: Institut za istrazivanja i projektovanja u privredi 2017-01-01
• Series: Istrazivanja i projektovanja za privredu
• Subjects: image classification; image model; statistical characteristics; fractal dimension of the scene; neural network without a teacher
• Online Access: https://scindeks-clanci.ceon.rs/data/pdf/1451-4117/2017/1451-41171703236I.pdf
• ISSN: 1451-4117; 1821-3197

Method for automated classification of ground forest inventory images based on the proposed mathematical model developed. The general model is represented by the statistical characteristics of images and fractal dimension of texture. Experimental means were determined minimally sufficient characteristics to solve the problem of robust classification. Neural network based on unsupervised self-organizing maps used as a classifier. Figures obtained discounts of the proposed approach on real digital images.