In　order　to　improve　the　effects　of　vision　based　car　detection,　a　vehicle　detection　method　based　on　image　entropy　is　proposed.　Firstly　a　Hough　transform　based　on　density　function　of　double　normal　distribution　on　theta-axis　is　proposed　for　lane　detection,　considering　complicated　and　structured　multi-Lane　driving　surroundings.　Then　preceding　vehicle　is　searched　and　located　using　a　method　with　image　entropy.　The　characteristics　of　image　entropy　on　vehicles　and　roads　are　fully　used　in　the　method,　and　have　a　strong　robustness　to　illumination.　Experiment　verifies　that　the　method　obtains　anticipative　effects　and　has　a　high　real-time　property　and　stability.