Plant species recognition based on bark patterns using novel Gabor filter banks

This paper presents a novel style of Gabor filter banks designed for plant species recognition using their bark texture features. In this paper, texture is modeled as multiple narrowband signals that are characterized by their central frequencies and normalized ratios of amplitudes. The normalized ratio of amplitudes is employed as an energy weight for combining narrowband signals. Based on this texture model, a set of texture features can be extracted from each kind of plant bark that is used to characterize the plant and to design the corresponding Gabor filter bank. A classifier is constructed by these Gabor filter banks. Plant recognition experiments on a small database of bark images have been conducted and the effectiveness of our approach is confirmed by the experimental results.