Classification of selected apple fruit varieties using naive bayes
Dr Miriti, Evans
MetadataShow full item record
Manual sorting of apple fruit varieties results to high cost, subjectivity, tediousness and inconsistency associated with human beings. A means for distinguishing apple varieties is needed and therefore, some reliable technique is needed to discriminate varieties rapidly and non-destructively. The main objective of this research was to investigate the applicability and performance of Naive Bayes algorithm in classification of apple fruit varieties. The software methodology involved image acquisition, pre-processing and segmentation, analysis and classification of apple varieties. Apple classification system prototype was built using MATLAB R2015 development platform environment. The results showed that the averaged values of the estimated accuracy, sensitivity, precision and specificity were 91%, 77%, 100% and 80% respectively. Through previous research works, the literature review identified MLP-Neural (Unay et al., 2006), fuzzy logic (Kavdir et al., 2003), principal components analysis (Bin et al., 2007) and neural networks (Ohali et al., 2011) as other technique which have been used previously to classify apple varieties. Benchmarking the performance of Naive Bayes technique against Principal Components Analysis, Fuzzy Logic and MLP-Neural classification technique showed that the Naive Bayes techniques performance was consistent with that of Principal Components Analysis, Fuzzy Logic and MLP-Neural with 91%, 90%, 89%, and 83% respectively in terms of accuracy. This study indicated that Naive Bayes has good potential for identification of apples varieties nondestructively and accurately. Keywords: Apple fruit, Sorting and Grading of Agricultural products, Image processing techniques, Naive Bayes Technique, Pattern Recognition, Classification.
University of Nairobi