The　key　and　difficult　point　for　video　targets　tracking　is　how　to　quickly　and　accurately　match　the　targets.　This　paper　presents　a　kind　of　moving　targets　tracking　method　based　on　multi-mode　fusion,　in　order　to　solve　the　tracking　algorithm　of　single　mode　can＇t　track　the　targets　accurately　under　the　complex　background.　This　algorithm　is　implemented　with　three　basic　patterns　including　Normalized　Cross　Correlation　(for　short,　NCC),　MeanShift　and　Kalman　filter,　together　with　the　Minimum　root-mean-square　error　criterion.　Computer　simulation　results　show　that　the　fused　algorithm　tracks　very　well　under　the　complex　background　and　it　has　the　robust　performance　of　approach.