Automatic recognition of plant leaves using serial combination of classifiers

Abstract

Plants are of great importance in human life, they are useful in many field such as industry, medicine, agriculture, etc. Plant identification is not a trivial task and presents challenges even for specialists. In this paper, we present an automatic leaf classification system based on a serial combination of two classifiers, namely: Linear discriminate analysis and Naïve Bayes. Our system is consisted of two stages, at the first stage, NB classifier attempts to determine, with a reject option, the class that a given sample is belonging to. If the confidence score yielded by NB does not exceed a certain threshold, then the sample will be passed through another classification task using LDA classifier. Our system has been evaluated using the well-known Swedish dataset. Experimental results indicated that the serial combination of the classifiers has shown better performance than those obtained using only one classifier.