According　to　color　and　geometric　properties　of　traffic　signs　in　our　country,　an　efficient　traffic　sign　recognition　system　applying　to　natural　scenes　is　proposed　in　this　paper.　In　this　system,　an　improved　image　segmentation　algorithm　based　on　RGB　model　is　implemented　on　segmenting　traffic　signs　in　natural　scenes.　Moreover,　two　level　neural　networks　are　used　to　classify　and　recognize　traffic　signs.　The　outline　and　invariable　moment　characteristics　are　used　as　the　input　characteristics　of　the　classification　neural　network　and　identification　neural　network,　respectively.　The　experimental　results　demonstrate　this　efficient　system　can　achieve　perfect　reorganization　results　to　traffic　signs　in　natural　sciences;　furthermore,　it＇s　robust　and　broad　applicability.