Moving　target　detection　is　widely　used　in　intelligent　video　surveillance,　multimedia　applications,　and　other　fields.　Compared　with　traditional　moving　target　detection,　moving　target　detection　of　UAV　video　images　has　the　same　ideas,　but　there　are　many　interference　factors.　Aiming　at　the　problem　of　moving　target　detection　of　UAV,　a　moving　target　detection　algorithm　based　on　accurate　motion　vector　compensation　is　proposed.　In　this　algorithm,　SURF　algorithm　is　used　to　obtain　the　matching　pair　of　motion　vectors,　and　RANSAC　algorithm　is　used　to　delete　the　matching　features　to　obtain　the　accurate　motion　vectors.At　the　same　time,　the　influence　of　the　target　motion　vector　on　the　global　motion　estimation　is　eliminated　by　the　adaptive　background　filtering　method.　By　establishing　the　affine　transformation　model　of　the　background　motion　vector,　the　background　motion　vector　is　substituted　into　the　model　for　parameter　estimation.After　the　motion　compensation　image　is　calculated,　the　moving　target　detection　results　are　obtained　by　inter-frame　difference.Experimental　results　show　that　the　algorithm　is　robust　and　can　accurately　detect　moving　targets　in　the　background　of　UAV　shooting.　©　2021　ACM.