With　the　increase　of　traffic　jams,　the　intelligent　vehicle　gains　more　and　more　attention.　They　exploit　exteroceptive　sensors　to　have　an　accurate　perception　of　the　surroundings　in　the　urban　environment,　like　a　multi-layer　laser　radar　or　camera.　This　paper　presents　an　occupancy　grid　framework　to　handle　uncertainty　sources　caused　by　laser　radar.　The　inverse　sensor　model　in　Cartesian　coordinate　is　used　to　transform　the　sensor　data　into　occupancy　grid　map.　Two　grids　are　used,　local　grid　map　which　transforms　the　current　sensor　data　and　global　grid　map　performing　the　temporal　integration　of　data　in　a　fixed　frame　based　on　Dempster＇s　rule　of　combination　and　PCR2.　The　conflict　information　is　used　to　detect　the　moving　objects,　the　expansion　algorithm,　erosion　algorithm　and　the　priority　marking　algorithm　based　on　regional　growing　are　used　for　the　cluster　analysis,　and　box　models　are　established　for　the　moving　objects.　The　outdoor　experimental　results　show　that　such　a　perception　strategy　can　steadily　and　accurately　detect　the　moving　objects.