The　article　presents　a　clustering　algorithm　based　on　feature-matching　of　Multilayer　Laserscanner　to　solve　false　alarm　caused　by　slopes　or　low　interfering　objects　in　the　complex　urban　environment.　The　experiments　show　this　method　can　filter　non-target　points　in　scene　and　access　to　the　real　targets　effectively.　To　solve　multi-target　recognition　on　the　road,　this　article　uses　a　method　which　combines　binary　tree　structure　and　high-precision　binary　classifier　based　on　Adaboost　to　transform　multi-target　classification　into　a　series　of　binary　ones.　The　experimental　results　show　that　both　target　detection　algorithm　and　recognition　algorithm　are　stable　and　work　well　in　detecting　and　recognizing　pedestrians　or　vehicles　on　the　road.