Effect on the Performance of a Support Vector Machine Based Machine Vision System with Dry and Wet Ore Sample Images in Classification and Grade Prediction

Abstract

The aim of the present study is to analysing the effect of water absorption on iron ore samples in the performances of SVM-based machine vision system. Two types of SVM-based machine vision system (classification and regression) were designed and developed, and performances were compared with dry and wet ore sample images. The images of the ore samples were captured in both the conditions (wet and dry) to examine the proposed model performance. A total of 280 image features were extracted and optimised using sequential forward floating selection (SFFS) algorithm for model development. The iron ore samples were collected from an Indian iron ore mine (Guamine), and image capturing system was fabricated in the laboratory for executing the proposed study. The results indicated that a different set of optimised features obtained for dry and wet sample images in both the models (classification and regression). Furthermore, the performance of both the models with dry sample images was found to be relatively better than the wet sample images.