The　variations　of　erosion　or　deposition　of　the　riverbed　caused　by　water　and　sediment　movement　are　common　phenomena　in　nature,　which　often　cause　practical　engineering　problems　such　as　river　sedimentation,　river　bed　deformation,　coastal　retreat,　and　reservoir　capacity　reduction.　These　are　closely　related　to　the　design,　construction　and　operation　of　hydraulic　projects.　Therefore,　it　is　of　great　significance　to　develop　accurate　and　efficient　methods　of　three　dimensional　riverbed　terrains　reconstruction　and　analyzing　the　volumes　of　erosion　or　deposition.　In　this　paper,　flume　experiments　with　bed　load　sediment　transport　were　conducted　to　study　the　river　bed　evolution　under　different　flow　conditions.　The　methods　of　Structure　from　Motion　(SFM)　were　introduced　in　details,　containing　the　shooting　method,　selection　of　GCPs,　the　establishment　of　river-channel　mesh　and　the　interpolation　method.　Based　on　the　SFM　method,　three-dimensional　terrain　reconstructions　of　bed　surface　were　carried　out　before　and　after　scouring　experiments　to　obtain　the　dense　point　clouds.　On　this　basis,　the　same　ground　control　points　(GCPs)　were　fixed　to　acquire　the　actual　three　dimensional　terrain　coordinates,　and　then　three　dimensional　point　cloud　coordinates　were　interpolated　in　the　river　bed　mesh　to　accurately　analyze　erosional　and　depositional　quantities　of　the　entire　bed.　The　results　show　that:　1)　the　volumes　with　or　without　the　presence　of　block　model　were　calculated　by　SFM　respectively　to　verify　that　the　relative　error　of　the　present　method　is　less　than　4%.　2)　The　variation　qualities　of　erosion　or　deposition　calculated　by　the　SFM　method　and　the　qualities　of　the　actual　weighing　were　compared　and　the　relative　error　of　the　present　method　is　less　than　5%.　3)　This　method　was　also　applied　to　the　large-scale　river　model　experiment　and　the　relative　error　is　less　than　10%.　The　results　above　show　that　present　method　can　be　applied　to　both　flume　experiment　and　larger　river　bed　physical　model　experiment,　to　reconstruct　the　three-dimensional　bed　topography　quickly　and　efficiently.　The　paper　offers　an　innovative　idea　of　studying　bed　evolution　in　hydraulic　model　experiments.　Copyright　©2020　Advanced　Engineering　Sciences.　All　rights　reserved.