With　the　background　of　robotic　mapping　in　unknown　environment,　this　paper　proposes　measures　to　improve　matching　rate　with　SIFT,　creates　a　novel　cellular　automata　model　of　unknow　environment　and　achieve　cellular　automata　navigation.　It　consists　of　three　parts.On　the　first,　the　paper　introduces　cellular　automata　environment　model.On　the　second,　scale-invariant　image　feature　is　used　to　extract　environment　feature　in　room,　and　also　It　proposed　an　approach　to　improve　error　matched　pixel.At　last,　due　to　sensor　model　for　binocular　stereo　vision　sensor,　local　3D　coordinates　of　features　was　converted　to　global　2D　coordinates　to　get　a　feature　map,through　validating　and　clustering　features　in　feature　map,　we　get　the　cellular　autamat　map,at　the　same　time　the　paper　finished　the　cellular　automata　navigation.　aiming　at　complex　environment　in　room.The　performance　of　the　proposed　algorithm　was　verified　by　experiment　in　home　environment　with　obstacle.　©　2011　IEEE.