Evaluation of image partitioning strategies for preserving spatial information of cross-sectional micrographs in automated wood recognition of Fagaceae

Abstract Although wood cross sections contain spatiotemporal information regarding tree growth, computer vision-based identification studies have traditionally favored disordered image representations that do not take such into account. This paper describes partitioning strategies preserve the spatial of cross-sectional images. Three are designed, namely grid based on pyramid matching and its variants, radial tangential partitioning, their recognition performance is evaluated for Fagaceae micrograph dataset. The achieve better than bag-of-features model constitutes underlying framework. Radial which a strategy preserving from pith to bark, further improves performance, especially radial-porous species. Pearson correlation autocorrelation coefficients produced radially partitioned sub-images potential be used as auxiliaries in construction multi-feature datasets. contribution found limited species unremarkable at genus level.