In　order　to　solve　the　problem　of　tracking　frame　drift　and　tracking　failure　in　traditional　Camshift　algorithm　under　color　interference　or　occlusion,　an　improved　multi　feature　fusion　method　for　Camshift　target　tracking　is　proposed.　Firstly,　the　color　feature　and　the　HOG　feature　of　the　target　are　extracted　respectively　and　fused　according　to　a　certain　weight　coefficient　to　establish　the　target　model;　secondly,　the　target　occlusion　is　judged　according　to　the　similarity　measurement.　When　the　target　is　occluded,　the　target　is　divided　into　four　parts,　and　the　color　feature　and　the　HOG　feature　of　the　target　are　extracted　respectively,　and　weighted　fusion　is　carried　out　to　establish　the　occluded　target　model;　finally,　a　linear　prediction　method　is　proposed　to　replace　the　traditional　Kalman　filtering　algorithm　to　complete　the　position　prediction,　and　substitute　the　predicted　position　results　into　the　Camshift　algorithm　for　tracking.　Experimental　results　show　that　the　algorithm　can　improve　the　robustness　and　meet　the　real-time　requirements.　©　2020　IEEE.