[{"data":1,"prerenderedAt":504},["ShallowReactive",2],{"method-surfelmeshing2020":3},{"method":4,"reference":58,"equipment":80,"figures":101,"results":102},{"id":5,"label":6,"shortName":7,"title":8,"year":9,"era":10,"cluster":11,"scope":12,"keyIdeaZh":13,"keyIdeaEn":14,"fulltextStatus":15,"publicationStatus":16,"recommendation":17,"constructionRelevance":18,"validationEnvironment":19,"strengths":22,"limitations":27,"sensors":33,"platform":35,"estimator":38,"association":39,"timeModel":40,"deskew":41,"loopClosure":42,"globalOptimization":43,"mapRepresentation":44,"prior":45,"outputGeometry":46,"compute":47,"codeUrl":48,"codeLicense":49,"relatedVersions":50},"surfelmeshing2020","Schöps et al., 2020","SurfelMeshing","SurfelMeshing: Online Surfel-Based Mesh Reconstruction",2020,"recent","C08","map_representation_or_reconstruction","SurfelMeshing 假設相機已校正且位姿由外部 SLAM 提供，不把深度融合進體素體積，而是融合成稠密面元（surfel）雲，再在背景非同步地對平滑後的面元做局部三角化，產生頂點即為面元的網格。作者在 ElasticFusion 式的面元重建上加入兩個去雜訊步驟：沿法向與鄰近面元的正則化，以及在觀測邊界漸進混合深度差以避免表面斷裂；並提出只在失效三角形附近重新三角化的增量演算法，使網格能隨迴圈閉合造成的面元變形快速更新。由於面元依輸入影像解析度建立，網格與色彩解析度會隨觀測距離調整，也能重建體素法難以保留的薄物體。","Online mesh reconstruction from a dense surfel cloud: ElasticFusion-style surfel fusion with new normal-direction regularization and observation-boundary blending, plus asynchronous local (re)triangulation so the mesh follows loop-closure deformations, adapts its resolution to the input and keeps thin objects; poses come from an external SLAM system.","full_text_reviewed","peer_reviewed_published","background","論文未涉及營建場景，評估以 TUM RGB-D、ICL-NUIM 與 CoRBS 室內小場景為主。它能在 SLAM 過程中持續輸出可隨迴圈閉合更新的網格，並保留薄物體與相機解析度的色彩，可作為室內構件即時網格預覽的參考；但前處理捨棄 3 m 以外的深度，大型室內空間或工地的適用性未經驗證（推論）。",[20,21],"public_benchmark","simulation",[23,24,25,26],"With ground-truth trajectories on ICL-NUIM (1 cm threshold) accuracy is higher than InfiniTAM and FastFusion on most sequences; with loop closures completeness is higher than ElasticFusion on kt0 to kt2 (Table 2)","Lowest mean curvature, i.e. smoothest surfaces, of all compared methods (Tables 1 and 2)","Incremental remeshing yields mesh quality close to meshing from scratch at far lower cost; meshing the final cloud from scratch takes about 5.6 s (Table 1; Fig. 10)","Reconstructs thin objects and colours at the camera's resolution (Sec. 5.3; Figs. 14 and 15)",[28,29,30,31,32],"Meshes are not guaranteed to be manifold and can contain small holes where smoothing is insufficient (Sec. 6; Fig. 18)","Large loop-closure deformations adopted from ElasticFusion may rip surfaces apart (Sec. 6)","Less complete than the TSDF methods when ground-truth poses are used; ElasticFusion is partly more accurate (Table 2; Sec. 5.2)","Loop closures cause short disruptions, about 680 ms for 3.1 million surfels; cost of the external SLAM is not included in timings (Sec. 5.4)","Depth beyond 3 m is dropped in preprocessing for the Kinect v1 data (Sec. 4)",[34],"RGB-D camera (mainly Microsoft Kinect v1 sequences of the TUM RGB-D benchmark; pre-registered CoRBS sequences; synthetic ICL-NUIM depth with simulated noise)",[36,37],"real RGB-D sequences from TUM RGB-D, CoRBS and ETH3D (capture platform not described in the paper; ETH3D poses from BAD SLAM)","simulation (ICL-NUIM synthetic living room)","No pose estimation of its own: calibrated camera and poses from an external SLAM system (ElasticFusion in the implementation; BAD SLAM poses for the ETH3D examples)","Projective association of each surfel with the pixel it projects to and the nearest neighbouring pixel; surfels classified as conflicting, occluded or supported using a depth uncertainty interval of plus or minus 5% of the measured depth and normal checks","discrete camera poses","not_applicable (RGB-D input)","Adopts ElasticFusion's loop-closure handling: the surfel cloud is deformed based on surfel timestamps, offsets are averaged among neighbours for 100 iterations, and affected mesh regions are remeshed; the public code excludes loop closure","none of its own; relies on the SLAM system's non-rigid surfel deformation","dense surfel cloud at input image resolution (position, normal, colour, confidence, radius, creation and update timestamps, denoised position, four neighbours) indexed by a lazily updated compressed octree; a triangle mesh whose vertices are the surfels","none","coloured triangle mesh with vertices at the surfels, updated online; not guaranteed manifold","GPU (CUDA 8.0) for surfel reconstruction and denoising at frame rate, CPU threads for asynchronous meshing and remeshing; tested on Intel Core i7 6700K with GeForce GTX 1080; average remeshing iteration 212 ms; a loop closure over 3.1 million surfels takes about 680 ms (Sec. 4, 5.4)","https:\u002F\u002Fgithub.com\u002Fpuzzlepaint\u002Fsurfelmeshing","BSD-3-Clause style licence (LICENSE file checked; copyright ETH Zurich, Thomas Schoeps)",[51,55],{"relation":52,"title":53,"doi_or_url":54},"accepted_manuscript","SurfelMeshing (arXiv v2, author's accepted version with IEEE notice)","https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.00729",{"relation":56,"title":57,"doi_or_url":48},"code_release","puzzlepaint\u002Fsurfelmeshing (loop closure functionality excluded)",{"id":5,"kind":59,"shortName":7,"title":8,"authors":60,"year":9,"venue":64,"venueType":65,"publisher":66,"volumeIssuePages":67,"doi":68,"arxivId":69,"url":70,"firstPublicDate":71,"publicationStatus":16,"metadataStatus":72,"fulltextStatus":15,"era":10,"classicReason":73,"codeUrl":48,"cluster":11,"topics":74,"mdpi":75,"verification":76,"label":6,"fulltextRoute":77,"versionRead":78,"addedByCensus":79},"method",[61,62,63],"Thomas Schöps","Torsten Sattler","Marc Pollefeys","IEEE Transactions on Pattern Analysis and Machine Intelligence","journal","IEEE","42(10):2494-2507","10.1109\u002Ftpami.2019.2947048","1810.00729","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1109\u002FTPAMI.2019.2947048","2018-10-01","metadata_verified","not_applicable",[11],false,"corrected","arXiv","arXiv 1810.00729v2 (2019-11-20), author's accepted version for IEEE TPAMI (header notes publisher changes before publication); IEEE version of record not compared",true,[81,88,95],{"category":82,"model":83,"canonical":83,"role":84,"dataset":85,"specs":86,"locator":87},"compute","PC with Intel Core i7 6700K and MSI Geforce GTX 1080 Gaming X 8G","compute for runtime",null,"surfel reconstruction and denoising in CUDA 8.0 on the GPU; meshing on CPU","Sec. 4; Sec. 5",{"category":89,"model":90,"canonical":90,"role":91,"dataset":92,"specs":93,"locator":94},"rgbd","Kinect v1","dataset sensor","TUM RGB-D","640x480 images; depth beyond 3 m dropped in preprocessing","Sec. 4; Sec. 5.3",{"category":89,"model":96,"canonical":97,"role":91,"dataset":98,"specs":99,"locator":100},"Kinect v2","Microsoft Kinect v2","CoRBS","pre-registered CoRBS sequences (sensor named in the title of reference [46])","Sec. 5; ref. [46]",[],{"totalRows":103,"groupCount":104,"groups":105,"others":498},30,5,[106,330,402,466],{"slug":107,"group":108,"sourceId":5,"sourceLabel":6,"table":109,"selfRows":110,"metrics":111,"seqs":123,"entrants":134,"cells":151,"outcomes":323,"locators":324,"hardware":326,"wordings":327,"notes":328},"surfelmeshing2020-table-2-ground-truth-trajectories","surfelmeshing2020:Table 2 (ground-truth trajectories)","Table 2 (ground-truth trajectories)",12,[112,117,119],{"label":113,"unit":114,"statistic":115,"alignment":116},"Accuracy [%] (share of reconstructed surfels within 1 cm of ground truth)","%","not_reported","SE3",{"label":118,"unit":114,"statistic":115,"alignment":116},"Completeness [%] (share of ground-truth points within 1 cm of the reconstruction)",{"label":120,"unit":121,"statistic":122,"alignment":116},"Curvature [0.01\u002Fm] (mean curvature, smoothness)","0.01 1\u002Fm","mean",[124,128,130,132],{"dataset":125,"sequence":126,"environment":127},"ICL-NUIM","kt0","synthetic indoor living room",{"dataset":125,"sequence":129,"environment":127},"kt1",{"dataset":125,"sequence":131,"environment":127},"kt2",{"dataset":125,"sequence":133,"environment":127},"kt3",[135,138,140,142,144,147,149],{"name":136,"methodId":137,"linkable":79,"proposed":75,"self":75},"InfiniTAM [29]","infinitam2015",{"name":139,"methodId":137,"linkable":79,"proposed":75,"self":75},"InfiniTAM [29] - smoothed",{"name":141,"methodId":85,"linkable":75,"proposed":75,"self":75},"FastFusion [27]",{"name":143,"methodId":85,"linkable":75,"proposed":75,"self":75},"FastFusion [27] - smoothed",{"name":145,"methodId":146,"linkable":79,"proposed":75,"self":75},"ElasticFusion [17]","elasticfusion2015",{"name":148,"methodId":146,"linkable":79,"proposed":75,"self":75},"ElasticFusion [17] - smoothed",{"name":150,"methodId":5,"linkable":79,"proposed":79,"self":79},"SurfelMeshing (Ours)",[152,156,159,162,165,168,170,173,175,177,179,181,183,185,187,189,191,193,195,197,199,201,203,205,207,209,211,213,215,217,219,221,223,225,227,228,230,232,234,236,238,240,242,244,246,248,249,251,253,255,257,259,261,262,264,266,268,270,272,274,276,278,280,282,284,286,288,290,292,294,296,298,300,302,304,306,308,309,311,313,315,317,319,321],[153,153,153,154,155,153,155,155,153],0,76.4,-1,[157,153,153,158,155,153,155,155,153],1,78.3,[160,153,153,161,155,153,155,155,153],2,85.5,[163,153,153,164,155,153,155,155,153],3,75.9,[166,153,153,167,155,153,155,155,153],4,96.2,[104,153,153,169,155,153,155,155,153],95.7,[171,153,153,172,155,153,155,155,153],6,93.5,[153,153,157,174,155,153,155,155,153],68.4,[157,153,157,176,155,153,155,155,153],68.3,[160,153,157,178,155,153,155,155,153],80.1,[163,153,157,180,155,153,155,155,153],78.8,[166,153,157,182,155,153,155,155,153],83.1,[104,153,157,184,155,153,155,155,153],82.1,[171,153,157,186,155,153,155,155,153],86.4,[153,153,160,188,155,153,155,155,153],56.6,[157,153,160,190,155,153,155,155,153],58.1,[160,153,160,192,155,153,155,155,153],64.2,[163,153,160,194,155,153,155,155,153],52.6,[166,153,160,196,155,153,155,155,153],97.1,[104,153,160,198,155,153,155,155,153],96.6,[171,153,160,200,155,153,155,155,153],69.5,[153,153,163,202,155,153,155,155,153],54.4,[157,153,163,204,155,153,155,155,153],58.3,[160,153,163,206,155,153,155,155,153],85.3,[163,153,163,208,155,153,155,155,153],72.5,[166,153,163,210,155,153,155,155,153],93.7,[104,153,163,212,155,153,155,155,153],92.2,[171,153,163,214,155,153,155,155,153],74,[153,157,153,216,155,153,155,155,153],53.5,[157,157,153,218,155,153,155,155,153],51.6,[160,157,153,220,155,153,155,155,153],54.7,[163,157,153,222,155,153,155,155,153],45.6,[166,157,153,224,155,153,155,155,153],38.8,[104,157,153,226,155,153,155,155,153],38.9,[171,157,153,222,155,153,155,155,153],[153,157,157,229,155,153,155,155,153],66.5,[157,157,157,231,155,153,155,155,153],62.8,[160,157,157,233,155,153,155,155,153],67.3,[163,157,157,235,155,153,155,155,153],63,[166,157,157,237,155,153,155,155,153],46,[104,157,157,239,155,153,155,155,153],45.3,[171,157,157,241,155,153,155,155,153],58.6,[153,157,160,243,155,153,155,155,153],46.7,[157,157,160,245,155,153,155,155,153],42.6,[160,157,160,247,155,153,155,155,153],48.4,[163,157,160,224,155,153,155,155,153],[166,157,160,250,155,153,155,155,153],22.8,[104,157,160,252,155,153,155,155,153],23.1,[171,157,160,254,155,153,155,155,153],30.8,[153,157,163,256,155,153,155,155,153],61.6,[157,157,163,258,155,153,155,155,153],59.9,[160,157,163,260,155,153,155,155,153],82.8,[163,157,163,229,155,153,155,155,153],[166,157,163,263,155,153,155,155,153],40,[104,157,163,265,155,153,155,155,153],40.4,[171,157,163,267,155,153,155,155,153],52.1,[153,160,153,269,155,153,155,155,153],2.48,[157,160,153,271,155,153,155,155,153],1.46,[160,160,153,273,155,153,155,155,153],0.99,[163,160,153,275,155,153,155,155,153],0.63,[166,160,153,277,155,153,155,155,153],0.24,[104,160,153,279,155,153,155,155,153],0.18,[171,160,153,281,155,153,155,155,153],0.15,[153,160,157,283,155,153,155,155,153],2.23,[157,160,157,285,155,153,155,155,153],0.93,[160,160,157,287,155,153,155,155,153],1.47,[163,160,157,289,155,153,155,155,153],0.87,[166,160,157,291,155,153,155,155,153],0.43,[104,160,157,293,155,153,155,155,153],0.35,[171,160,157,295,155,153,155,155,153],0.17,[153,160,160,297,155,153,155,155,153],4.71,[157,160,160,299,155,153,155,155,153],1.71,[160,160,160,301,155,153,155,155,153],1.68,[163,160,160,303,155,153,155,155,153],1.2,[166,160,160,305,155,153,155,155,153],0.34,[104,160,160,307,155,153,155,155,153],0.28,[171,160,160,279,155,153,155,155,153],[153,160,163,310,155,153,155,155,153],3.69,[157,160,163,312,155,153,155,155,153],1.34,[160,160,163,314,155,153,155,155,153],1.33,[163,160,163,316,155,153,155,155,153],0.92,[166,160,163,318,155,153,155,155,153],0.47,[104,160,163,320,155,153,155,155,153],0.41,[171,160,163,322,155,153,155,155,153],0.32,[],[325],"Table 2",[],[],[329],"ICL-NUIM living room with simulated depth noise; ground-truth trajectories used and loop-closure handling disabled for all methods; reconstructions aligned to the ground-truth model with point-to-plane ICP; evaluation threshold 1 cm; 'smoothed' = same bilateral filter as SurfelMeshing preprocessing",{"slug":331,"group":332,"sourceId":5,"sourceLabel":6,"table":333,"selfRows":334,"metrics":335,"seqs":339,"entrants":343,"cells":347,"outcomes":396,"locators":397,"hardware":398,"wordings":399,"notes":400},"surfelmeshing2020-table-2-loop-closure-trajectories","surfelmeshing2020:Table 2 (loop-closure trajectories)","Table 2 (loop-closure trajectories)",9,[336,337,338],{"label":113,"unit":114,"statistic":115,"alignment":116},{"label":118,"unit":114,"statistic":115,"alignment":116},{"label":120,"unit":121,"statistic":122,"alignment":116},[340,341,342],{"dataset":125,"sequence":126,"environment":127},{"dataset":125,"sequence":129,"environment":127},{"dataset":125,"sequence":131,"environment":127},[344,345,346],{"name":145,"methodId":146,"linkable":79,"proposed":75,"self":75},{"name":148,"methodId":146,"linkable":79,"proposed":75,"self":75},{"name":150,"methodId":5,"linkable":79,"proposed":79,"self":79},[348,350,352,354,356,358,360,362,364,366,368,370,372,374,376,377,379,380,382,384,386,388,390,392,393,394,395],[153,153,153,349,155,153,155,155,153],95.8,[157,153,153,351,155,153,155,155,153],96.8,[160,153,153,353,155,153,155,155,153],87.2,[153,153,157,355,155,153,155,155,153],64.9,[157,153,157,357,155,153,155,155,153],64.4,[160,153,157,359,155,153,155,155,153],58.5,[153,153,160,361,155,153,155,155,153],26.9,[157,153,160,363,155,153,155,155,153],26.8,[160,153,160,365,155,153,155,155,153],35,[153,157,153,367,155,153,155,155,153],40.6,[157,157,153,369,155,153,155,155,153],41.8,[160,157,153,371,155,153,155,155,153],44,[153,157,157,373,155,153,155,155,153],34.6,[157,157,157,375,155,153,155,155,153],34.5,[160,157,157,369,155,153,155,155,153],[153,157,160,378,155,153,155,155,153],5.8,[157,157,160,378,155,153,155,155,153],[160,157,160,381,155,153,155,155,153],15.9,[153,160,153,383,155,153,155,155,153],0.86,[157,160,153,385,155,153,155,155,153],0.37,[160,160,153,387,155,153,155,155,153],0.22,[153,160,157,389,155,153,155,155,153],0.39,[157,160,157,391,155,153,155,155,153],0.3,[160,160,157,281,155,153,155,155,153],[153,160,160,389,155,153,155,155,153],[157,160,160,293,155,153,155,155,153],[160,160,160,391,155,153,155,155,153],[],[325],[],[],[401],"ICL-NUIM living room with simulated depth noise; trajectories estimated with ElasticFusion including loop closures (kt3 omitted because ElasticFusion failed); aligned with point-to-plane ICP; threshold 1 cm; InfiniTAM and FastFusion cannot handle loop closures and have no values",{"slug":403,"group":404,"sourceId":5,"sourceLabel":6,"table":405,"selfRows":171,"metrics":406,"seqs":411,"entrants":419,"cells":424,"outcomes":460,"locators":461,"hardware":462,"wordings":463,"notes":464},"surfelmeshing2020-table-1","surfelmeshing2020:Table 1","Table 1",[407,409],{"label":408,"unit":114,"statistic":115,"alignment":115},"bdry: amount of vertices on a mesh boundary",{"label":410,"unit":121,"statistic":122,"alignment":115},"crv: mean curvature in 0.01\u002Fm",[412,415,417],{"dataset":92,"sequence":413,"environment":414},"fr1\u002Fdesk","indoor office scenes, Kinect v1 (carrying mode not stated in the paper)",{"dataset":92,"sequence":416,"environment":414},"fr1\u002Fxyz",{"dataset":92,"sequence":418,"environment":414},"fr3\u002Foffice",[420,422,423],{"name":421,"methodId":5,"linkable":79,"proposed":79,"self":79},"Ours (regularization, blending and remeshing enabled)",{"name":136,"methodId":137,"linkable":79,"proposed":75,"self":75},{"name":141,"methodId":85,"linkable":75,"proposed":75,"self":75},[425,427,429,431,433,435,437,439,441,443,445,447,449,450,452,454,456,458],[153,153,153,426,155,153,155,155,153],2.3,[157,153,153,428,155,153,155,155,153],9.7,[160,153,153,430,155,153,155,155,153],17.7,[153,153,157,432,155,153,155,155,153],1.7,[157,153,157,434,155,153,155,155,153],11.1,[160,153,157,436,155,153,155,155,153],15.6,[153,153,160,438,155,153,155,155,153],1.3,[157,153,160,440,155,153,155,155,153],5.5,[160,153,160,442,155,153,155,155,153],18.3,[153,157,153,444,155,153,155,155,153],0.6,[157,157,153,446,155,153,155,155,153],7.4,[160,157,153,448,155,153,155,155,153],8.4,[153,157,157,391,155,153,155,155,153],[157,157,157,451,155,153,155,155,153],3.9,[160,157,157,453,155,153,155,155,153],7.2,[153,157,160,455,155,153,155,155,153],0.5,[157,157,160,457,155,153,155,155,153],2.6,[160,157,160,459,155,153,155,155,153],6.6,[],[405],[],[],[465],"Mesh quality on TUM RGB-D reconstructions (Kinect v1 data; the trajectories used for Table 1 are not specified); truncated extract of Table 1: full method and the two volumetric baselines, boundary-vertex share and mean curvature only",{"slug":467,"group":468,"sourceId":5,"sourceLabel":6,"table":469,"selfRows":160,"metrics":470,"seqs":476,"entrants":480,"cells":482,"outcomes":487,"locators":489,"hardware":492,"wordings":494,"notes":495},"surfelmeshing2020-text-sec-5-4","surfelmeshing2020:Text Sec.5.4","Text Sec.5.4",[471,474],{"label":472,"unit":473,"statistic":122,"alignment":115},"average time per remeshing iteration","ms",{"label":475,"unit":473,"statistic":115,"alignment":115},"handling a loop closure for 3.1 million surfels",[477],{"dataset":92,"sequence":478,"environment":479},"fr3\u002Flong_office_household","indoor office, Kinect v1 (carrying mode not stated in the paper)",[481],{"name":150,"methodId":5,"linkable":79,"proposed":79,"self":79},[483,485],[153,153,153,484,155,153,153,155,153],212,[153,157,153,486,153,157,153,155,157],680,[488],"other: approximate value ('ca. 680ms', Sec. 5.4)",[490,491],"Sec. 5.4; Fig. 19","Sec. 5.4",[493],"Intel Core i7 6700K + GeForce GTX 1080 (external SLAM cost excluded)",[],[496,497],"Performance on the TUM fr3\u002Flong office household sequence at 640x480; remeshing statistics averaged over iterations","Time to handle one loop closure (deformation and smoothing) for 3.1 million surfels",[499],{"group":500,"slug":501,"sourceLabel":6,"table":502,"selfRows":157,"datasets":503},"surfelmeshing2020:Text Sec.5.1","surfelmeshing2020-text-sec-5-1","Text Sec.5.1",[92],1790510664456]