As　the　result　of　the　requirements　of　accuracy　and　speediness　of　indoor　localization　algorithm,　we　improved　the　performance　of　corner　detection　algorithm.　In　this　paper,　the　theory　of　multi-scale　is　introduced　into　the　classical　harris　algorithm,　and　detects　local　maximum　points　at　each　scale　level.　This　method　might　overcome　the　drawback　that　the　single-scale　harris　detector　usually　leads　to　either　missing　significant　corners　or　detecting　false　corners　due　to　noise,　and　it　not　only　maintains　the　advantages　of　traditional　harris　corner　which　is　invariant　to　the　changes　of　intensity　and　camera　pose　but　also　can　be　used　in　multi-scale.　Experimental　results　demonstrate　the　effectiveness　of　the　proposed　algorithm.　©　2010　IEEE.