Phenotypic analysis of Arabidopsis Thaliana root plant with improved feature extraction and combining classifiers approach

• Published in: 2012 16th CSI International Symposium on Artificial Intelligence and Signal Processing pp. 452 - 457
• Main Authors: Farhidzadeh, H., Hashemi, S. N., Masoudnia, S.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.05.2012
• Subjects: Acceleration; Arabidopsis Thaliana; Artificial neural networks; Correlation; Feature extraction; gated network; Genetics; mixture of experts; negative correlation learning; Organisms; sliding window; Training
• ISBN: 9781467314787; 1467314781
• DOI: 10.1109/AISP.2012.6313790

In this paper, we present a modified feature extraction and an improved combining classifiers method to analyse, model and classify plants growth process. Plants growth is a significant issue in different aspects in biology. Arabidopsis Thaliana is a plant that is very much interesting, because its genetic structure has some similarities with that of human. Its reaction against genetic mutations is an important subject for scientists that can divide it as mutated type plant or not mutated one (wild type). Due to its fast root growth, morphological changes are not observable. Making time series of consecutive pictures of root growth process and by using pattern recognition techniques, it is controversial to classify and determine plants type. For this purpose, we enhance feature extraction process by intervening root growth velocity and acceleration. Moreover, moving a sliding window on dataset helps to improve classification rate. We employ an improved version of negative correlation learning (NCL) method in which the capability of gating network, as the combining part of Mixture of Expert (ME) method, is used to combine the base Neural Networks (NNs) in the NCL ensemble method. Experimental results show its usefulness in comparison with the other classical methods such as SVM and NCL.