Intelligent　sensors　serve　as　crucial　elements　in　the　realm　of　smart　car　mobility　solutions　and　urban　sensing　technology.　This　article　presents　a　novel　automotive　environment　perception　system　that　uses　a　binocular　vision　sensor.　The　binocular　camera　is　used　to　capture　images　and　obtain　cloud　points　for　obstacle　perception　and　environment　positioning.　The　proposed　system　is　built　on　a　low-power　embedded　platform　but　maintains　a　high　perception　performance.　It　can　accurately　identify　and　locate　obstacles,　such　as　vehicles　and　pedestrians.　The　complete　system　is　comprehensively　described,　encompassing　the　hardware　structure,　software　architecture,　and　algorithm　program.　Furthermore,　the　process　of　the　obstacle　detection　algorithm,　which　relies　on　disparity　space　and　deep　learning　(DL),　is　thoroughly　presented.　The　feasibility　of　the　fast　stereo-matching　algorithm　is　demonstrated　theoretically　and　validated　through　experimental　verification.　Extensive　experimental　results　indicate　that　the　system　is　capable　of　delivering　reliable　and　precise　real-time　environmental　perception　for　intelligent　vehicles.　Consequently,　the　system　can　be　readily　implemented　in　commercial　real-time　intelligent　driving　applications.　As　a　pertinent　research　in　urban　sensing　applications,　this　system　holds　promise　as　a　viable　solution　for　enhancing　smart　mobility.　©　2001-2012　IEEE.