期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2016
卷号：7
期号：1
页码：433-440
出版社：TechScience Publications
摘要：Citrus industry contributes a major part innation’s growth, but there has been a decrease in production ofgood quality citrus fruits, due to improper cultivation, lack ofmaintenance, very high post harvest losses in handling andprocessing, manual inspection, lack of knowledge ofpreservation and quick quality evaluation techniques.Unrelated features, along with repetitive features, severelyaffect the accuracy of the learning machines. So, feature subsetcollection should be able to identify and remove as much of theirrelevant and redundant information as possible. A featureselection algorithm may be evaluated from both the efficiencyand effectiveness. The efficiency relates to the time spend tofind only relevant features from collection, the effectivenessconcerns to the quality of the required features. Based on theseconditions, an improved clustering-based feature selectionalgorithm is experimented. The improved/efficient clusteringmethods are implemented in two stages. In the first stage,features are divided into clusters by using graph-theoreticclustering algorithms. In the second stage, the importantfeature that is strongly related to target classes is selected fromeach cluster to form a subset of features. The efficiency of theeffective clustering algorithm are evaluated through anempirical study. The specific objectives implemented toaccomplish is: collect images from citrus leaves of threecommon citrus diseases, and normal leaves; determine colorco-occurrence method texture features for each image in thedataset; Apply effective Clustering and Classification toretrieve feature data models. In this paper, determine theeffective clustering/ classification accuracies using aperformance measure for feature extraction in citrus fruits andleaves.
