In　order　to　enhance　the　real-time　and　stability　of　lane　detection　based　on　machine　vision,　a　method　for　line　detection　based　on　combined　road　model　is　proposed.　After　image　classification　according　to　illumination,　the　image　is　processed　by　different　algorithms　during　image　pretreatment.　For　straight　line　and　curve　line　in　the　vicinity,　improved　Hough　Transform　(HT)　is　adopted　for　line　detection　and　tracking.　Both　of　the　road　boundaries　are　fitted　using　Catmull-Rom　Splines　based　on　control　points　search　algorithm　for　curve　line　in　the　distance.　For　various　kinds　of　lanes　on　most　structural　road,　experiment　results　indicate　that　the　method　has　good　robustness　and　stability.