rock images

Rock segmentation is a crucial task in geotechnical engineering, planetary exploration, and geological surveying. Traditional approaches rely heavily on manual intervention and expert domain knowledge, making them time-consuming, subjective, and prone to inconsistencies. With the advent of unsupervised machine learning techniques, there has been a growing interest in automating rock identification and segmentation using clustering- based algorithms. This paper introduces an end-to-end pipeline that leverages K-Means clustering to automate rock segmentation from RGB imagery. Implemented in Python using OpenCV and scikit-learn, the pipeline processes 196 rock images and segments them based on color and textural features. The system also provides visualizations in the form of color-coded clusters and mineral heatmaps to enhance interpretability. We performed a detailed literature review of existing segmentation methodologies, analyze current limitations, and present a gap analysis. The experimental results show that while K-Means provides a fast and interpretable baseline, future research should aim to combine clustering with explainable AI and adaptive models. This work contributes to making field surveys safer, faster, and more efficient by automating one of the most labor-intensive aspects of rock analysis.