Effect of near-infrared images with Rgb images for identifying Garnet mineral in Pulmuddai beach Sri Lanka

Abstract

Pulmuddai beach sand is one of the foremost mineral manufacturing sources in Sri Lanka. Pulmuddai Seaside is ironic in, Rutile, Ilmenite, Garnet, and Zircon. It is the prime mineral processing company in Sri Lanka. It's an excessive requirement to discover high mineral zones for manufacturing. In manufacturing, the percentage of the mineral is calculated using a visual examination through a microscope. It is a manual process which is time- consuming. The study identified the effect of near-infrared images to differentiate Garnet mineral from sand by using image processing methods. Near-Infrared (NIR) is a portion of the electromagnetic spectrum unseen to the naked eye. The frequency ranges between 650nm to 950nm. Through a strong beam of infrared rays, any object can be envisioned with high precision. The NIR response is free from color dependency and can be simply understood by an observer. In this study, 100 visible light RGB (Red, Green, Blue) images and 100 NIR images of sand were captured. It was captured in a controlled light environment. The RGB color images and NIR images were examined separately to recognize the Garnet mineral from the sample and to compute the Garnet percentage. Machine learning classification (support vector machine) purpose variance, contrast, kurtosis, skewness, mean, median, min, max, range, standard deviation, and correlation were mined from sand grains images. Then RGB and Red, Green, Blue, and near-infrared (RGBNIR) color models were used to train the machine learning model. The RGBNIR color model increased the separation accuracy by 23% compared to the RGB-only model. The accuracy might be enlarged by presenting more images to the machine learning procedure.