In　recent　years,　with　the　progress　of　image　processing　technology,　the　research　on　Simultaneous　Localization　and　Mapping　(SLAM)　has　gradually　become　a　hot　topic　in　the　field　of　computer　vision.　Visual　SLAM　is　mainly　divided　into　front　end　and　back　end.　In　this　study,　aiming　at　the　problem　of　low　efficiency　of　feature　point　matching,　an　improved　fast　feature　approximation　nearest　neighbor　feature　matching　method　is　proposed,　which　can　obtain　higher　precision　image　matching　relationship　in　a　short　time.　At　the　same　time,　an　algorithm　of　Cluster　Sampling　Consensus　(CLUSAC)　algorithm　is　proposed,　which　reduces　the　probability　of　mismatching　and　makes　camera　pose　estimation　more　accurate.　Finally,　a　comparative　experiment　of　the　algorithm　is　performed.　The　results　show　that　the　improved　feature　matching　algorithm　and　mismatching　elimination　algorithm　can　effectively　improve　the　accuracy　of　feature　matching　and　help　to　provide　accurate　and　effective　initial　values　for　the　back　end.　©　2019　IEEE.