The　fingerprint-based　localization　technique　has　become　a　focus　in　indoor　localization　since　it　has　the　merits　of　strong　anti-interference　ability,　good　network　stability,　high　quality　data　transmission　and　wide　applicability.　For　fingerprint-based　localization,　however,　offline　site　survey　is　usually　time-consuming　and　labor-intensive.　Therefore,　offline　site　survey　reduction　becomes　an　important　issue　for　fingerprint-based　indoor　localization.　In　this　paper,　we　first　model　the　fingerprints　of　all　reference　points　(RPs)　as　a　low-tubal-rank　tensor　and　then　propose　a　two-pass　adaptive　sampling　scheme　where　under　a　given　budget　of　samples,　a　fixed　proportion　of　samples　are　firstly　allocated　uniformly　at　random,　and　then　the　remaining　samples　are　assigned　according　to　the　leverage　scores　of　the　underlying　tensor.　Afterward,　we　apply　a　tensor　completion　algorithm　such　as　ADMM,　t-TNN　and　TenALS　to　solve　the　weighted　nuclear　norm　minimization　problem　as　above.　The　advantage　is　that　the　scheme　has　ability　to　identify　the　highly　informative　RPs　that　are　conductive　to　fingerprint　map　reconstruction　to　mitigate　the　sample　complexity　for　offline　site　survey.　Simulation　results　show　that　the　proposed　adaptive　sampling　scheme　can　significantly　improve　the　performance　in　terms　of　both　reconstruction　error　and　localization　accuracy　compared　with　the　random　uniform　sampling　method.