An　automatic　registration　algorithm　for　the　scattered　point　clouds　is　proposed　in　this　paper.　The　curvature　of　each　point　is　estimated　by　using　the　quadratic　surface　fitting　method.　And　then,　according　to　the　change　of　curvature　the　feature　points　that　have　the　maximum　local　changes　are　extracted.　The　initial　match　points　are　acquired　by　computing　the　Hausdorff　distance　of　curvature,　and　the　feature　of　the　surface　circumference　shape　is　used　on　the　initial　match　points　to　obtain　the　accurate　match　points.　Finally,　the　rotation　and　translation　matrix　are　estimated　by　the　quaternion,　and　iterative　solution　is　used　to　improve　the　registration　accuracy.　So,　the　automatic　match　of　the　scattered　point　cloud　data　is　completed.　Experimental　results　show　that　the　algorithm　is　effective.