Tree leaf feature extraction and recognition based on geometric features and Haar wavelet theory

Abstract

In the grim situation of wood shortage, efficient utilize forest resources and rational use of wood have an important significance. Different kinds of trees have different use-value, so it is very important to identify the species of trees. Different species of trees have their own leaf characteristics. In this study, we proposed a novel feature extraction method based on geometric features and Haar wavelet, which can achieve the tree leaves feature rapid extraction. Extracting the geometrical features of leaves, at the same time, make Haar wavelet triple decomposition to the leaf image, calculating the leaves statistical characteristics like energy, entropy and mean values etc. Finally realize the recognition of tree species. The experimental results show that geometric features and statistical characteristics have significantly different, these differences can effectively identify the types of tree by using the classic adaboost threshold classifier, and the method is effective and practicable.