<p>Characterizing rock hardness and abrasivity is essential for drill-bit design. However, intact cores are often unobtainable in complex formations and in extraterrestrial settings, precluding conventional laboratory determination of these indices. In this context, we develop models that relate hardness and abrasivity to mineral composition and microstructural parameters, and examine how these factors connect to mechanical behavior. The dataset comprises 64 rock samples with quartz and feldspar contents and grain sizes, together with hardness and abrasivity. The key advance is a fully automated workflow for microstructural quantification. Gray-level co-occurrence matrix (GLCM) texture features are extracted from thin-section images and then interpreted qualitatively to tie these statistics to grain size, surface condition, cleavage development, and fabric, thereby improving interpretability. Before feature extraction, histogram equalization enhances image texture. Correlation analysis reveals strong cross-orientation redundancy among texture metrics and significant correlations among quartz and feldspar contents and grain sizes, which supports feature reduction. For hardness prediction, support vector regression with a Gaussian kernel performs best. Quartz content is the dominant control, and adding physically meaningful texture measures such as contrast and energy significantly improves accuracy, and highlights the nonlinear microstructural effects. For abrasivity, a linear kernel is more suitable, reflecting a stronger linear dependence on mineral composition; quartz grain size is the primary control, followed by quartz and feldspar contents. Texture features add little to abrasivity prediction. Together, these results unite interpretable automated texture metrics with mineralogical information, clarify the differing controls on hardness and abrasivity, and offer guidance for bit design and engineering applications.</p>