In　China,　managing　mixed　traffic　flow　consisting　of　mobile,　bicycle,　and　pedestrians　at　signalized　intersection　is　one　of　the　key　measures　to　ensure　a　good　traffic　operation.　Focusing　on　the　bicycle　groups　at　signalized　intersection,　the　video　capture　technology　is　used　to　collect　their　passing　characters　at　two　intersections　respectively　in　Beijing　and　Tianjin　city.　Through　the　K-means　cluster　analysis　method,　three　criterions　are　selected　to　divide　the　passing　stage:　the　average　speed,　width　and　length　of　bicycle　groups　at　the　two　signalized　intersections.　The　passing　stage　can　be　divided　into　three　stages;　gradually　spreading　stage,　the　expanding　stage,　and　the　gradually　shrinking　stage.　©2011　by　Science　Press.