Author(s): ANIL KANNUR, ASHA KANNUR, VIJAY S RAJPUROHIT
Journal: International Journal of Machine Intelligence
ISSN 0975-2927
Volume: 3;
Issue: 2;
Start page: 62;
Date: 2011;
VIEW PDF
DOWNLOAD PDF 
Original page
Keywords: Classification | Grading | Watershed | Extraction | Seeds | Neural | Elman’s | Cascade-Forward | Feed-Forward
ABSTRACT
This paper describes a different neural network model for classification and grading of bulk seeds samples usingdifferent artificial neural network models. Algorithms are developed to acquire and process color images of bulk seedssamples. Different seeds like Groundnut, Jowar, Wheat, Rice, Metagi, Red gram, Bengal gram, and Lentils etc. areconsidered for the study. The developed algorithms are used to extract over 11 (9 color, area and equidiameter) features, 18(color only) features and 20 (18 color and 2 boundary) features. The area and equidiameter features are extracted from thewatershed segmentation. Different types of Neural Network based classifier is used to identify the unknown seeds samples.The classification is carried out using different types of features sets, viz., color, area and equidiameter. Classificationaccuracies of over 85% are obtained for all the seeds types using all the three feature sets. And also different neural networkgives different accuracies and time period taken for training all the three feature sets.
Journal: International Journal of Machine Intelligence
ISSN 0975-2927
Volume: 3;
Issue: 2;
Start page: 62;
Date: 2011;
VIEW PDF
DOWNLOAD PDF Original pageKeywords: Classification | Grading | Watershed | Extraction | Seeds | Neural | Elman’s | Cascade-Forward | Feed-Forward
ABSTRACT
This paper describes a different neural network model for classification and grading of bulk seeds samples usingdifferent artificial neural network models. Algorithms are developed to acquire and process color images of bulk seedssamples. Different seeds like Groundnut, Jowar, Wheat, Rice, Metagi, Red gram, Bengal gram, and Lentils etc. areconsidered for the study. The developed algorithms are used to extract over 11 (9 color, area and equidiameter) features, 18(color only) features and 20 (18 color and 2 boundary) features. The area and equidiameter features are extracted from thewatershed segmentation. Different types of Neural Network based classifier is used to identify the unknown seeds samples.The classification is carried out using different types of features sets, viz., color, area and equidiameter. Classificationaccuracies of over 85% are obtained for all the seeds types using all the three feature sets. And also different neural networkgives different accuracies and time period taken for training all the three feature sets.
