Bark recognition using novel rotationally invariant multispectral textural features

0506602 - ÚTIA 2020 RIV NL eng J - Journal Article
Remeš, Václav - Haindl, Michal
Bark recognition using novel rotationally invariant multispectral textural features.
Pattern Recognition Letters. Roč. 125, č. 1 (2019), s. 612-617. ISSN 0167-8655. E-ISSN 1872-7344
R&D Projects: GA ČR(CZ) GA19-12340S
Institutional support: RVO:67985556
Keywords : Bark recognition * Tree taxonomy clasification * Spiral Markov random field model * textural feature
OECD category: Communication engineering and systems
Impact factor: 3.255, year: 2019
Method of publishing: Limited access
http://library.utia.cas.cz/separaty/2019/RO/haindl-0506602.pdf https://www.sciencedirect.com/science/article/pii/S0167865519301886

We present novel rotationally invariant fully multispectral Markovian textural features applied for the efficient tree bark recognition. These textural features are derived from the novel descriptive multispectral spiral wide-sense Markov model. Unlike the alternative bark recognition methods based on various gray-scale discriminative textural descriptions, we benefit from fully descriptive color, rotationally invariant bark texture representation. The proposed methods significantly outperform the state-of-the-art bark recognition approaches regarding classification accuracy. Both our classifiers outperform convolutional neural network ResNet even on the largest public bark database BarkNet which contains 23 000 high-resolution images from 23 different tree species.
Permanent Link: http://hdl.handle.net/11104/0297826