In　this　paper,　the　expansion　of　feature　points　of　the　linear　scale　space　is　transformed　into　the　classification　of　multi-scale　data　set　within　the　same　scale,　which　belongs　to　the　classification　of　scale　invariant　non-equilibrium.The　paper　presents　a　sample　approach　based　on　kernel　learning　to　solve　classification　on　imbalance　dataset　by　Support　Vector　Machine　(SVM).　The　method　first　preprocesses　the　data　by　oversampling　the　minority　class　in　kernel　space,　and　then　the　pre-images　of　the　synthetic　samples　are　found　based　on　a　distance　relation　between　kernel　space　and　input　space.　Finally,　these　pre-images　are　appended　to　the　original　dataset　to　train　a　SVM.　As　a　result,　the　inconsistency　which　is　brought　about　by　processing　samples　in　different　spaces　is　overcome.　On　the　other　hand,　the　sampling　strategies　of　the　method　not　only　can　decrease　imbalanced　rate　of　training　dataset,　but　also　can　enlarge　convex　hull　of　the　minority　class.　Consequently,　the　problem　of　the　boundary　skew　can　be　amended　more　effectively.　Experiments　on　real　dataset　indicate　that　the　generalization　performance　of　the　result　classifier　is　improved　and　the　algorithm　can　work　well　on　expanding　the　feature　points　stably　for　a　certain　scale.　©　2012　ACADEMY　PUBLISHER.