Ultrasonic　testing　plays　a　crucial　role　in　detecting　early　structural　damage　and　identifying　micro-defects,　particularly　in　processes　like　additive　manufacturing　and　welding.　The　full　matrix　capture　(FMC)　method,　leveraging　laser　ultrasound　technology,　excels　in　imaging　sub-millimeter　micro　defects.　However,　its　extensive　data　acquisition　time　hinders　real-time　imaging.　To　address　this,　a　selection　matrix　capture　approach　is　adopted　to　reduce　data　collection　and　enhance　detection　speed.　Specifically,　a　multi-parameter　genetic　algorithm　(MPGA)　is　proposed　to　optimize　sparse　array　layouts.　This　optimization　is　based　on　theoretical　detection　sensitivity　means　and　standard　deviations,　evaluating　array　layout　quality.　The　imaging　method　combined　multi-scale　principal　component　analysis　with　phase　weighting　techniques.　Experiments　on　sub-millimeter　defects,　including　side　drilling　holes　(SDH),　blind　holes　(BH),　and　spherical　holes　(SH),　were　conducted.　Results　showed　that,　compared　to　random　and　uniform　sparsity,　the　genetic　algorithm　optimized　sparse　array　provided　superior　imaging　as　sparsity　decreased.　Effective　defect　detection　was　achieved　with　only　5　%-20　%　of　full　matrix　data.