Identification of dockage in paddy using multiclass SVM

Abstract

A colour machine vision system was used for identification of dockage including organic and inorganic impurities, varietal admixture (paddy variety KPL 1) and grain admixture (whole undehulled black gram) in paddy. The image acquisition system consisted of a CMOS camera, LED ring lighting and a sample platform. A dockage identification model was designed for identification and classification of dockage types using multi-class support vector machine (SVM) classifier. The algorithm considered 5 morphological and 9 colour features from the acquired images and results showed an overall classification accuracy of 90.3%. The mean classification accuracies for paddy, organic and inorganic impurities, varietal admixture and grain admixture were 85%, 83.2%, 93%, 98% and 93.6%, respectively. This approach can be used as a rapid, non-destructive quality evaluation technique for paddy based on visual attributes.