Classification of images of wheat, ryegrass and brome grass species at early growth stages using principal component analysis

Abstract

Wheat is one of the most important crops in Australia, and the identification of young plants is an important step towards developing an automated system for monitoring crop establishment and also for differentiating crop from weeds. In this paper, a framework to differentiate early narrow-leaf wheat from two common weeds from their digital images is developed. A combination of colour, texture and shape features is used. These features are reduced to three descriptors using Principal Component Analysis. The three components provide an effective and significant means for distinguishing the three grasses. Further analysis enables threshold levels to be set for the discrimination of the plant species. The PCA model was evaluated on an independent data set of plants and the results show accuracy of 88% and 85% in the differentiation of ryegrass and brome grass from wheat, respectively. The outcomes of this study can be integrated into new knowledge in developing computer vision systems used in automated weed management. © 2011 Golzarian and Frick; licensee BioMed Central Ltd.