Obstacle　detection　has　been　one　of　the　core　issues　of　research　in　the　field　of　intelligent　vehicle　visual　navigation　system,　auto　auxiliary　driving,etc.　This　paper　proposes　a　new　detection　method　based　on　monocular　vision,　which　is　used　to　detect　any　type　of　obstacle　in　the　rear　of　the　vehicle　when　the　vehicle　in　low-speed　reversing.　First,　we　use　the　Forward-Backward　error(FB　error)　on　the　original　image　to　select　the　easy　tracking　and　reliable　feature　points.　Second,　using　these　feature　points,the　feature　points　belonging　to　ground　are　selected　in　top-view　images　converted　from　original　images.　Based　on　road　assumption　of　plane,　vehicle　motion　parameters　can　be　estimated　and　we　compensate　for　the　rotation　and　translation　of　the　vehicle　using　the　estimated　parameters.　Finally,　using　the　different　movement　between　the　obstacle　and　the　road,　we　exclude　noise　points　and　extract　obstacle　region　in　the　difference　image　based　on　motion　compensation.　The　experimental　results　show　that　the　method　proposed　in　this　paper　can　detect　any　type　of　obstacle.　©　2013　TCCT,　CAA.