Performance Analysis of Cereals Grains using Neural Network & Multi Support Vector Machine

Abstract

Objectives: To classify cereal grain using a multi-support vector machine and artificial neural network for better accuracy. To build a system for cereals grains classification with the use of image processing techniques. Methods: Using a CCD camera, the method starts with image acquisition. To acquire images Grayscale conversion, noise reduction, binarization, edge detection, and morphological operations are applied. Using the edge detection technique edge of the objects is predictable. The watershed algorithm is used for the segmentation of touching and overlapping cereals kernels. Local Binary Pattern (LBP) texture feature and color features extracted from segmented images. For image classification, the features extraction method is used. Findings: We have incorporated various parameters like shape, size, length, width, major axis length, and minor axis lengths on different cereals like rice, barley, millet, sorghum, wheat, and millet. There are a total of 96 images of the data set are used to train or test the model. Out of that 70% are training and 30% are testing. Improvements: In the proposed MSVM technique, we have achieved 89.7% accuracy and in the ANN technique the accuracy is 92.3% which is higher than the conventional SVM technique. Novelty: The proposed technique is based on the Multi Support Vector Machine (MSVM) and Artificial Neural Network (ANN). We have compared the MSVM and ANN with the SVM technique.