Feature selection and classification improvement of Kinnow using SVM classifier. (December 2022)

• Record Type: Journal Article
• Authors: Singh, Sukhpreet; Malik, Kamal
• Is Part Of: Measurement. Volume 24(2022)
• Publication Date: 2022-12
• Subjects: Classification -- Feature extraction -- Feature selection -- Kinnow -- Normalization; Detectors -- Periodicals; Measurement -- Periodicals; 530.7
• DOI: 10.1016/j.measen.2022.100518
• Languages: English
• ISSNs: 2665-9174

Abstract

Abstract: An approach to enhance the classification of the kinnow is proposed in this paper. The fruit images are captured in the proposed approach, and texture, color, shape, and size features are extracted and merged to generate a dataset. To cope with outliers and dominant features, the Pareto normalization method is used. A hybrid feature selection approach that combines neighborhood component analysis and ReliefF methods to select the optimal features is proposed to eliminate redundant and irrelevant features. The dataset is classified using the SVM machine learning algorithm following the feature selection. Utilizing the SVM classifier, the proposed approach chooses 54.20% of the features with an accuracy of 94.67%. This proves that the proposed approach is efficient and can be used for the classification of the other fruits. Highlights: Categories of grades of Kinnow are elaborated for better clarity. EM algorithm is added under image segmentation section. Citation is added for better understanding of segmentation and classification.