Moving　target　detection　in　the　dynamic　background,　such　as　optical　flow　detect,　is　an　expensive　method,　which　is　hard　to　satisfy　its　real　time　requirement　in　the　project.　In　this　paper,　a　new　method　is　proposed,　which　synthesizes　Lucas-Kanade,　Hierarchical　clustering　and　secondary　clustering　algorithms　together　that　can　rapidly　detect　moving　targets.　Since　Hierarchical　clustering　is　a　high　time　exhausting　algorithm,　in　order　to　reduce　the　time　consuming,　parts　of　the　clustering　algorithm　are　replaced　by　interframe　calculation　and　secondary　clustering.　The　experiments　show　that　this　method　is　accurate　and　real-time　in　detecting　the　moving　target　in　the　moving　background.　©　2012　AICIT.