Through　the　installation　of　a　monocular　vision　sensor　on　a　target　and　shooting　sequence　images　of　the　floor,　combined　with　the　characteristics　of　the　sequence　images,　an　indoor　mobile　visual　localization　algorithm　based　on　Harris-SIFT　is　proposed.　The　algorithm　is　first　to　establish　first-order　DOG　scale　space　of　sequence　images　of　the　floor,　to　extract　feature　points　in　each　layer　images　of　DOG　scale　space　by　using　Harris　operator,　then　to　select　scale-invariant　feature　points　to　calculate　each　feature　point　descriptor　and　match,　and　thus　to　depict　the　trajectory　of　the　movement　of　the　target　to　achieve　positioning.　Using　several　different　indoor　sequence　images　under　the　different　environment　respectively,　the　algorithm　is　verified,　positioning　experimental　results　show　the　effectiveness　of　the　algorithm.