Evaluating the Use of a Contextual Information Extraction Technique to Identify Mineralized Zones in a Semi-Arid Environment from Aster Satellite Data

Abstract

Identification of regions of mineralization by traditional techniques where spectral information of pixel alone is applied during classification, either at pixel or sub-pixel level, is usually accompanied by some level of un-satisfaction. Impulse noises that are usually experienced in digital images from sudden sharp disturbances in the signal degrade the output. This effect often referred to as the salt and pepper noise could further cause information loss, and change the colour of an RGB image. The use of filters (median and morphological) has not totally eliminated the effects. Object-based methods came in with higher filter smoothers to make it better yet, there is potential limitation because of possible negative impact of under segmentation. The errors of under-segmentation cannot be adjusted within a unit of features, which apparently affect the potential accuracy of the entire classification. Thus, this study evaluates the contribution of the contextual information to reduce the effects of noise in the data for effective mineral identification. Rule-based technique was applied for information extraction from a threshold values derived from band ratio (BR) transformation operations on ASTER data. The result indicates clay has the highest mineral density of 47% in the study area, with silicate having the least (3%), among others. This study provides a robust test for contextual cues as anticipated to be most effective and shall contribute towards reducing environmental impacts and protecting biodiversity which is one of the major aspects of sustainable development in relation to mining and mineral processing