Classification of Cane Sugar Based on Physical Characteristics Using SVM

The document presented is all about classification of commercially available cane sugars. A variety of image processing algorithms are implemented. In this study, the researchers identified ten different classes of sugars and trained the system to the physical characteristics of each sugar. The main objective of this study is to classify cane sugar based on physical characteristics using SVM. Specifically, (1) to create a hardware that will categorize sugar based from color and size, (2) to implement different image processing technique using MATLAB, (3) to test and verify different crystalized white and brown sugar if the system is classifying correctly. The sugar will be classified into ten different kinds which include granulated sugar, caster sugar, cane sugar, sanding sugar, demerara sugar, turbinado sugar, muscovado sugar, light brown sugar, pearl sugar, and confectioners' sugar. The physical characteristics are known to be the color in RGB value and size in pixels. Darkfield Illumination will be implemented in to enhance and image captured by the camera. The training consists of 30 distinct sets of values for every sugar. This ensures that the project created will be able to successfully classify sugars. As a result, the device has a total accuracy of 86.97%.