In　order　to　solve　the　problems　of　costly　instruments　and　small　detection　area　in　visibility　detection,　a　traffic　visibility　estimation　algorithm　was　proposed　by　combing　light　transmission　model　in　fog　with　camera　calibration　geometrical　optics　model.　The　algorithm　calculated　the　distance　from　point　in　road　area　to　the　camera　by　calibrating　internal　and　external　parameters　dynamically,　obtaining　the　atmospheric　extinction　coefficient　and　estimating　the　traffic　visibility　using　scene　transmittance.　Firstly,　interesting　area　was　searched　by　activity　map,　and　the　fog　weather　was　recognized　by　fitting　curve　of　area　average　pixels　whether　meet　the　edge　spread　function　or　not.　Secondly,　transmittance　value　of　each　point　in　traffic　scene　was　calculated　by　dark　channel　prior,　and　four　group　points　of　transmission　difference　maximum　were　selected　to　calibrate　the　internal　parameters　of　the　camera.　Thirdly,　vanishing　point　and　boundary　of　the　road　were　extracted　to　calibrate　the　camera　external　parameters　dynamically.　Finally,　traffic　visibility　was　estimated　by　using　the　distance　from　the　point　in　road　area　to　the　camera　and　corresponding　scene　transmittance.　In　this　paper,　visibility　estimation　results　are　compared　with　corresponding　data　got　by　manual　and　physical　equipment.　We　also　verify　the　effectiveness　and　real-time　of　this　proposed　method.　©,　2015,　Science　Press.　All　right　reserved.