In　this　paper,　we　study　the　problem　of　moving　target　detection　in　video　sequences.　An　improved　frame　difference　target　detection　algorithm　is　proposed　on　the　basis　of　the　self　updating　average　background　model.　By　using　numerical　difference　and　logical　operations　between　the　average　background　and　the　current　frame,　the　empty　areas　which　are　derived　from　traditional　frame　difference　method　can　be　extracted.　For　eliminating　the　isolated　noise　caused　by　background　perturbation,　this　paper　puts　forward　the　neighborhood　two　values　discriminant　filtering　method.　Then,　a　compensation　scheme　based　on　the　information　of　the　empty　areas　together　with　the　results　of　frame　difference　method　is　employed　to　make　up　the　regions　of　moving　object.　Finally,　simple　comparisons　between　our　algorithm　and　other　methods　are　given.　The　experimental　results　show　that　the　integrity　of　moving　object　extracted　by　our　algorithm　has　been　greatly　improved　than　the　frame　difference　method.　Besides,　in　disturbance　scene,　it　has　better　performance　of　disturbance　rejection　than　the　mixture　of　Gaussians　model　on　the　premise　of　guaranteeing　the　integrity　of　the　moving　target.　©　2017　IEEE.