Read the procedure carefully: the population is first partitioned into relevant homogeneous subgroups, and then random selection happens within every subgroup. The decisive clue is sampling from within each subgroup, which is the signature of stratified random sampling, where the subgroups are the strata. This differs from cluster sampling, where the population is broken into clusters and only some entire clusters are chosen, with members of unselected clusters never sampled. It also differs from simple random sampling, which ignores subgroups altogether, and from systematic sampling, which selects every $k^{th}$ element from an ordered list. Since each stratum contributes a random sample, the technique is stratified sampling.
\[\boxed{\text{Stratified sampling}}\]