The　unstable　quality　of　3D　printing　products　is　one　of　the　obstacles　that　hinder　the　popular　application　of　3D　printing　technology.　In　this　paper,　we　propose　a　novel　mask　projection　3D　printing　scheme　with　visual　surveillance.　We　introduce　a　vision　system　to　monitor　the　gray　level　variation,　so　that　two　kinds　of　problems　could　be　detected　in　time:　1.　the　printed　objects　are　stuck　on　the　bottom　of　the　tank.　2.　the　printed　objects　are　stretched　out　and　part　of　objects　are　stuck　on　the　bottom　of　the　tank.　The　normal　and　abnormal　printing　procedures　correspond　to　different　gray　level　variation　curves,　which　can　be　discriminated　by　classify　algorithms.　By　the　computing　capability　of　our　computer,　we　could　monitor　gray　level　variation　curves　at　470　points　in　real　time.　If　95　%　curves　are　normal,　the　layer　is　thought　successfully　printed,　the　next　layer　will　be　printed　continuously.　Otherwise,　this　layer　is　abnormal,　and　the　printing　will　be　stopped.　The　experiment　show　that　the　proposed　scheme　can　stop　the　abnormal　printing　in　time.　it　can　improve　the　production　rate　of　the　finished　products　and　reduce　the　material　waste.