Linear Discriminant Analysis for Apples Fruit Variety Based on Color Feature Extraction

Choosing apples in large quantities depends on human eyesight which is still objective, so it needs a sorting process using technology so that it is feasible as industrial raw material. One technology that can be applied is to implement digital image processing. Through image processing, segmentation and classification techniques are deemed appropriate to carry out the sorting process. In this study, apples were classified by type, namely Braeburn Apples, Granny Smith, Pink Lady, Red Yellow, Golden and Red Delicious. In this study, 500 training images and 400 testing images of apples were used. The classification algorithm chosen is Linear Discriminant Analysis (LDA). LDA was chosen because it has characteristics in distinguishing two or several objects, can be used to reduce dimensions, and can divide data into several groups. Based on these advantages, LDA will be used in classifying apples based on skin color. To distinguish the characteristics of the fruit color, RGB
$\mathrm{L}^{*}\mathrm{a}^{*}\mathrm{b}$
color extraction is used. The application is made using the Matlab 2017b programming language. The research that has been done has produced an accuracy of 98%.