LIBS is getting particular attention in geological applications and is used throughout the mining value chain, from prospecting to output quality control. The preliminary screening of potential mining sites and geology prospecting, in general, demand an analytical technique with high throughput. Laser-Induced Breakdown Spectroscopy (LIBS) proved to be a vital alternative, yet certain issues still remain to be solved. The main concern is the supervised classification of minerals that runs automatically without prior knowledge of the sample. Such a task demands establishing vast data libraries, building the classification methodology, and obtaining its non-biased prediction. We aim to demonstrate methodological steps for automated mineralogy of large-scale LIBS imaging (over 1Mpix hyperspectral cubes) on a series of samples from mining sites. To prove the concept and validate results, we incorporate correlative imaging of LIBS with energy-dispersive X-ray microprobe (SEM-EDX).