For　striking　enemy　armored　vehicles　from　the　air　or　on　the　ground,　it　is　a　necessary　to　timely　mastery　enemy　tanks　and　armored　vehicles　by　UAVs　on　the　battlefield.　The　traditional　target　detection　algorithms　have　shortcomings　such　as　insufficient　accuracy　and　slow　computing　speed,　which　make　the　monitoring　of　target　armored　vehicles　still　have　many　problems.　In　this　paper,　through　a　large　amount　of　image　data　containing　armored　vehicles,　the　improved　YOLOv5s　algorithm　is　used　in　the　field　of　deep　learning　for　target　detection　of　armored　vehicles.　Firstly,　the　SimAM　attention　mechanism　is　combined　with　the　C3　module　in　the　Backbone　network　to　improve　the　network＇s　ability　to　process　image　fine　textures.　Subsequently,　the　up-sampling　method　of　the　original　network　is　replaced　with　the　CARAFE　module.　The　experimental　results　show　that　the　detection　precision　and　recall　of　the　improved　network　model　are　increased　by　1.823%　and　6.89%,　respectively,　which　effectively　improves　the　accuracy　of　UAVs　in　recognizing　tank-armored　vehicles.