Identifikasi Varietas Padi Menggunakan Pengolahan Citra Digital dan Analisis Diskriminan

Abstract

Identification of the authenticity of variety is important in order to ensure the purly of rice variety. Variety identification based on visual traits is a common technique, but it is difficult to be applied when ones encounter almost similar varieties. An alternative technique is using the image processing analysis combined with the discriminant analysis. The research objectives were: (1) to develop the geometrical, shape and textural description of rice varieties (Ciherang, Inpari 10 and Inpari 13) generated based on the image process properties, (2) to determine selective geometry, shape and texture variables, and (3) to identify rice variety using discriminant analysis. Digital image of rice grains was captured by a scanner, then was quantified by image process application in order to generate geometry, shape and texture data. Discriminant model was built using 5 of 17 input variables that have strong discriminant power. Discriminant model represented 85.9% of the data variability. The model accuracy based on the cross validation method for Ciherang, Inpari 10 dan Inpari 13 varieties was 53.6%, 52.8% dan 76.0%, respectively. Image process technology and discriminant analysis has a potential as a technique for identifying rice variety based on the seed physical criteria. Further works need to be done in order to increase the accuracy of the model.