Topic　detection　recognizes　the　new　topic　in　a　stream　of　news　stories.　This　task　becomes　the　hotspot　research　direction　in　the　field　of　natural　language　processing　in　recent　years.　In　this　paper,　the　effective　features　in　the　story　document　are　extracted,　and　the　vector　centered　model　is　taken　to　represent　the　topic　and　the　story.　The　algorithm　of　incremental　clustering　is　carried　out　for　topic　detection,　which　will　identify　the　emergence　of　new　events　and　also　merge　the　story　to　the　corresponding　topic　cluster.　Finally,　we　give　the　statistical　data　analysis　and　achieve　the　performance　of　75%　F　value　on　the　test　set.　Topic　detection　will　also　provide　the　efficient　guidance　for　judging　the　hot　spots　in　the　web.