[{"data":1,"prerenderedAt":363},["ShallowReactive",2],{"method-chisel2015":3},{"method":4,"reference":55,"equipment":78,"figures":106,"results":107},{"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":37,"estimator":39,"association":40,"timeModel":41,"deskew":42,"loopClosure":43,"globalOptimization":44,"mapRepresentation":45,"prior":46,"outputGeometry":47,"compute":48,"codeUrl":49,"codeLicense":50,"relatedVersions":51},"chisel2015","Klingensmith et al., 2015","CHISEL","Chisel: Real Time Large Scale 3D Reconstruction Onboard a Mobile Device using Spatially Hashed Signed Distance Fields",2015,"classic","C08","odometry_with_local_mapping","CHISEL 在 Google Tango 手機與平板上，只用行動裝置的 CPU 即時建立房屋尺度（300 平方公尺以上）的 TSDF 稠密重建，不使用 GPU 通用運算。作者採用 Nießner 等人的空間雜湊兩層結構，把 16×16×16 體素的區塊（chunk）依視錐裁剪結果動態配置，未被更新的區塊即回收，只處理含有表面的空間以節省記憶體與運算。針對 Tango 深度感測器雜訊大、更新率只有 3 至 6 Hz 的問題，加入依雜訊模型調整的動態截斷距離與空間雕刻（space carving）。定位以裝置內建的視覺慣性里程計為輸入，再以掃描對 TSDF 模型的 ICP 修正短距漂移；網格則以增量 marching cubes 依區塊延遲產生，系統本身沒有迴圈閉合。","CPU-only house-scale TSDF reconstruction on Google Tango devices: spatially hashed 16^3-voxel chunks allocated by frustum culling and garbage-collected, dynamic truncation and space carving from a trained depth-noise model, visual-inertial odometry corrected by scan-to-TSDF ICP, and lazy per-chunk incremental meshing; no loop closure.","full_text_reviewed","peer_reviewed_published","background","論文未在施工現場測試。實例包括整層辦公大樓、約 175 公尺走廊、公寓與夜間戶外場景，皆以手持 Tango 裝置即時掃描（Fig. 1、7、8）。在行動裝置上即時建立 2 至 3 公分解析度的樓層模型，符合低成本手持室內掃描的需求；但系統會累積漂移（走廊末端約 5 公尺），且論文沒有以幾何真值評估精度，用於量測用途前需另以控制點校核（推論）。",[20,21],"public_benchmark","completed_building",[23,24,25,26],"Real-time house-scale reconstruction at 2 to 3 cm voxel resolution entirely on a mobile CPU (Abstract; Sec. IV-B)","Spatial hashing needed about a tenth of the fixed-grid memory as the explored space grew, never more than 47 MB for a 15 m hallway (Sec. IV-D, Fig. 5a)","Per-frame meshing and update of 102 ms and 128 ms with spatial hashing versus 2067 ms and 3769 ms for a 256^3 fixed grid (Table II)","Space carving strongly reduces noise artifacts around object silhouettes and removes briefly seen moving objects (Sec. III-E, IV-C, Fig. 9)",[28,29,30,31,32],"No global consistency; the map drifts over time, about 5 m at the end of a roughly 175 m corridor with VIO and dense alignment only (Sec. III-J, Sec. V, Fig. 7)","Resolution of 2 to 3 cm is much coarser than GPU TSDF systems that reach sub-centimetre voxels (Sec. V)","Tango depth arrives at only 3 to 6 Hz, too slow for depth-only tracking (Sec. III-J)","Projection mapping at 3 cm causes aliasing on surfaces nearly parallel to the viewing axis; the constant weighting approximation degrades surfaces in noisy areas (Sec. III-B, IV-C)","Quantitative results cover timing and memory only; no surface accuracy against ground truth is reported (Sec. IV)",[34,35,36],"Google Tango 'Peanut' phone: projective depth sensor (6 Hz), 120 deg wide-angle tracking camera (60 Hz), 4 MP colour camera (30 Hz), six-axis gyroscope and accelerometer","Google Tango 'Yellowstone' tablet: projective depth sensor (3 Hz), same tracking camera, 4 MP colour camera (30 Hz)","Kinect RGB-D data of the Freiburg (TUM) benchmark for memory experiments",[38],"handheld (Tango phone and tablet; office building floor, 175 m corridor, apartment, outdoor night scene)","Onboard visual-inertial odometry (EKF fusing wide-angle camera and inertial data with 2D feature tracking at 60 Hz; the paper refers to Kottas et al. and the MSCKF for details) with sparse keypoint mapping used as a black box, incrementally corrected by scan-to-model ICP against the TSDF (residual = TSDF value at each transformed point, gradient by central differences); corrective transforms accumulated per scan","Implicit point-to-TSDF association via the signed distance value and its gradient (first-order projection onto the zero level set)","discrete poses","not_applicable (depth camera)","none online; Fig. 7 shows a corridor corrected only after offline bundle adjustment","none online (offline bundle adjustment used only for the Fig. 7 corridor illustration)","dynamic spatially hashed TSDF: chunks of 16 x 16 x 16 voxels in a hash map, allocated on frustum intersection and garbage-collected when not updated; each voxel stores a 16-bit fixed-point SDF and 16-bit weight plus 8-bit RGB and colour weight; dynamic truncation from a trained depth-noise model; space carving","none","per-chunk triangle meshes from incremental marching cubes (lazy, asynchronous), coloured by trilinear interpolation; map savable to disk; 2 to 3 cm voxels on the devices","CPU only on the mobile device (no general-purpose GPU computing); single-scan fusion 62 to 200 ms on the Tango tablet and 14 to 58 ms on a desktop depending on the fusion mode (Fig. 9e)","https:\u002F\u002Fgithub.com\u002Fpersonalrobotics\u002FOpenChisel","MIT (stated in open_chisel\u002Fpackage.xml and source headers; no LICENSE file in the repository)",[52],{"relation":53,"title":54,"doi_or_url":49},"code_release","personalrobotics\u002FOpenChisel (open-source ROS reference implementation; the paper omits the link for double-blind review)",{"id":5,"kind":56,"shortName":7,"title":8,"authors":57,"year":9,"venue":62,"venueType":63,"publisher":64,"volumeIssuePages":65,"doi":66,"arxivId":67,"url":68,"firstPublicDate":69,"publicationStatus":16,"metadataStatus":70,"fulltextStatus":15,"era":10,"classicReason":71,"codeUrl":49,"cluster":11,"topics":72,"mdpi":73,"verification":74,"label":6,"fulltextRoute":75,"versionRead":76,"addedByCensus":77},"method",[58,59,60,61],"Matthew Klingensmith","Ivan Dryanovski","Siddhartha S. Srinivasa","Jizhong Xiao","Robotics: Science and Systems XI (RSS 2015)","conference","Robotics: Science and Systems Foundation","not_reported (paper 40)","10.15607\u002Frss.2015.xi.040",null,"https:\u002F\u002Fwww.roboticsproceedings.org\u002Frss11\u002Fp40.pdf","2015-07-13","metadata_verified","principle reused: CPU-only spatially hashed TSDF with frustum-culled chunk allocation, noise-model-based dynamic truncation and space carving on a handheld depth device; OpenChisel became a common CPU TSDF baseline (compared in densesurfelmapping2019 and flashfusion2018).",[11],false,"corrected","publisher OA","RSS XI online proceedings PDF (version of record, paper p40)",true,[79,85,88,94,101],{"category":80,"model":81,"canonical":81,"role":82,"dataset":67,"specs":83,"locator":84},"mobile_scanner_device","Google Tango 'Yellowstone' tablet","method input","4 GB RAM, quad-core CPU, Nvidia Tegra K1 graphics, 120 deg FOV tracking camera at 60 Hz, projective depth sensor at 3 Hz, 4 megapixel colour sensor at 30 Hz","Sec. IV-A",{"category":80,"model":86,"canonical":86,"role":82,"dataset":67,"specs":87,"locator":84},"Google Tango 'Peanut' mobile phone","2 GB RAM, quad-core CPU, six-axis gyroscope and accelerometer, 120 deg wide-angle tracking camera at 60 Hz, projective depth sensor at 6 Hz, 4 megapixel colour sensor at 30 Hz",{"category":89,"model":90,"canonical":90,"role":91,"dataset":67,"specs":92,"locator":93},"compute","desktop machine (model not reported)","compute for runtime","used only for the fusion-time comparison in Fig. 9e","Sec. IV-C; Fig. 9e",{"category":95,"model":96,"canonical":96,"role":97,"dataset":98,"specs":99,"locator":100},"rgbd","Kinect","dataset sensor","Freiburg (TUM) RGB-D benchmark","sensor loops around a central desk in a roughly 12 by 12 m room; the paper writes only 'Kinect', manufacturer inferred from the product name","Sec. IV-D",{"category":102,"model":103,"canonical":103,"role":104,"dataset":98,"specs":105,"locator":100},"other","motion capture system (model not reported)","reference or ground truth","ground-truth poses of the Freiburg dataset",[],{"totalRows":108,"groupCount":109,"groups":110,"others":350},50,6,[111,218,274,310],{"slug":112,"group":113,"sourceId":5,"sourceLabel":6,"table":114,"selfRows":115,"metrics":116,"seqs":124,"entrants":129,"cells":146,"outcomes":211,"locators":212,"hardware":213,"wordings":215,"notes":216},"chisel2015-fig-9e","chisel2015:Fig. 9e","Fig. 9e",32,[117,122],{"label":118,"unit":119,"statistic":120,"alignment":121},"single scan fusion time (ms.)","ms","mean","not_reported",{"label":118,"unit":119,"statistic":123,"alignment":121},"std",[125],{"dataset":126,"sequence":127,"environment":128},"authors' Room (apartment) dataset","Room","indoor apartment, handheld Tango",[130,132,134,136,138,140,142,144],{"name":131,"methodId":5,"linkable":77,"proposed":77,"self":77},"CHISEL: Raycast, no colour, no carving",{"name":133,"methodId":5,"linkable":77,"proposed":77,"self":77},"CHISEL: Raycast, colour, no carving",{"name":135,"methodId":5,"linkable":77,"proposed":77,"self":77},"CHISEL: Raycast, no colour, carving",{"name":137,"methodId":5,"linkable":77,"proposed":77,"self":77},"CHISEL: Raycast, colour, carving",{"name":139,"methodId":5,"linkable":77,"proposed":77,"self":77},"CHISEL: Projection mapping, no colour, no carving",{"name":141,"methodId":5,"linkable":77,"proposed":77,"self":77},"CHISEL: Projection mapping, colour, no carving",{"name":143,"methodId":5,"linkable":77,"proposed":77,"self":77},"CHISEL: Projection mapping, no colour, carving",{"name":145,"methodId":5,"linkable":77,"proposed":77,"self":77},"CHISEL: Projection mapping, colour, carving",[147,151,154,156,158,160,162,164,166,168,170,172,174,177,178,180,182,185,186,188,190,192,193,195,197,199,200,202,204,206,207,209],[148,148,148,149,150,148,148,150,148],0,14,-1,[148,152,148,153,150,148,148,150,148],1,2,[148,148,148,155,150,148,152,150,148],62,[148,152,148,157,150,148,152,150,148],13,[152,148,148,159,150,148,148,150,148],20,[152,152,148,161,150,148,148,150,148],5,[152,148,148,163,150,148,152,150,148],80,[152,152,148,165,150,148,152,150,148],16,[153,148,148,167,150,148,148,150,148],53,[153,152,148,169,150,148,148,150,148],10,[153,148,148,171,150,148,152,150,148],184,[153,152,148,173,150,148,152,150,148],40,[175,148,148,176,150,148,148,150,148],3,58,[175,152,148,165,150,148,148,150,148],[175,148,148,179,150,148,152,150,148],200,[175,152,148,181,150,148,152,150,148],37,[183,148,148,184,150,148,148,150,148],4,33,[183,152,148,161,150,148,148,150,148],[183,148,148,187,150,148,152,150,148],106,[183,152,148,189,150,148,152,150,148],22,[161,148,148,191,150,148,148,150,148],39,[161,152,148,161,150,148,148,150,148],[161,148,148,194,150,148,152,150,148],125,[161,152,148,196,150,148,152,150,148],23,[109,148,148,198,150,148,148,150,148],34,[109,152,148,183,150,148,148,150,148],[109,148,148,201,150,148,152,150,148],116,[109,152,148,203,150,148,152,150,148],19,[205,148,148,173,150,148,148,150,148],7,[205,152,148,161,150,148,148,150,148],[205,148,148,208,150,148,152,150,148],128,[205,152,148,210,150,148,152,150,148],24,[],[114],[214,81],"desktop (model not reported)",[],[217],"Time to fuse a single depth scan on the 'Room' (apartment, Fig. 1b) dataset for each fusion mode; mean and standard deviation printed as 'mean +- std'",{"slug":219,"group":220,"sourceId":221,"sourceLabel":222,"table":223,"selfRows":109,"metrics":224,"seqs":229,"entrants":238,"cells":242,"outcomes":267,"locators":268,"hardware":269,"wordings":271,"notes":272},"flashfusion2018-table-iv","flashfusion2018:Table IV","flashfusion2018","Han & Fang, 2018","Table IV",[225,227],{"label":226,"unit":119,"statistic":121,"alignment":121},"TSDF Fusion (ms)",{"label":228,"unit":119,"statistic":121,"alignment":121},"Mesh Extraction (ms)",[230,234,236],{"dataset":231,"sequence":232,"environment":233},"TUM RGB-D","fr3\u002Foffice (5mm voxels)","indoor office desk loop",{"dataset":231,"sequence":235,"environment":233},"fr3\u002Foffice (10mm voxels)",{"dataset":231,"sequence":237,"environment":233},"fr3\u002Foffice (20mm voxels)",[239,240],{"name":7,"methodId":5,"linkable":77,"proposed":73,"self":77},{"name":241,"methodId":221,"linkable":77,"proposed":77,"self":73},"FlashFusion",[243,245,247,249,251,253,255,257,259,261,263,265],[148,148,148,244,150,148,148,150,148],483,[148,152,148,246,150,148,148,150,148],1518,[152,148,148,248,150,148,148,150,148],3.6,[152,152,148,250,150,148,148,150,148],38.4,[148,148,152,252,150,148,148,150,148],86,[148,152,152,254,150,148,148,150,148],312,[152,148,152,256,150,148,148,150,148],1.1,[152,152,152,258,150,148,148,150,148],19.7,[148,148,153,260,150,148,148,150,148],15,[148,152,153,262,150,148,148,150,148],70,[152,148,153,264,150,148,148,150,148],0.7,[152,152,153,266,150,148,148,150,148],6.5,[],[223],[270],"Intel Core i7 7700 @3.6 GHz (CPU only)",[],[273],"Efficiency comparison between CPU-based CHISEL and FlashFusion on TUM fr3\u002Foffice at three voxel resolutions; time per operation in ms",{"slug":275,"group":276,"sourceId":5,"sourceLabel":6,"table":277,"selfRows":183,"metrics":278,"seqs":288,"entrants":292,"cells":295,"outcomes":304,"locators":305,"hardware":306,"wordings":307,"notes":308},"chisel2015-table-i","chisel2015:Table I","Table I",[279,282,284,286],{"label":280,"unit":281,"statistic":121,"alignment":46},"% of Bounding Box (Unknown Culled)","%",{"label":283,"unit":281,"statistic":121,"alignment":46},"% of Bounding Box (Unknown)",{"label":285,"unit":281,"statistic":121,"alignment":46},"% of Bounding Box (Outside)",{"label":287,"unit":281,"statistic":121,"alignment":46},"% of Bounding Box (Inside)",[289],{"dataset":98,"sequence":290,"environment":291},"Freiburg 5m (desk loop)","indoor room about 12 by 12 m, Kinect loop around a central desk (Freiburg dataset; carrying mode not stated)",[293],{"name":294,"methodId":5,"linkable":77,"proposed":77,"self":77},"CHISEL spatially hashed TSDF",[296,298,300,302],[148,148,148,297,150,148,150,150,148],77,[148,152,148,299,150,148,150,150,148],15.5,[148,153,148,301,150,148,150,150,148],4.1,[148,175,148,303,150,148,150,150,148],3.4,[],[277],[],[],[309],"Voxel statistics for the Freiburg 5 m depth-frustum reconstruction; share of the bounding box per voxel class (culled voxels are not stored)",{"slug":311,"group":312,"sourceId":5,"sourceLabel":6,"table":313,"selfRows":183,"metrics":314,"seqs":321,"entrants":323,"cells":328,"outcomes":343,"locators":344,"hardware":345,"wordings":347,"notes":348},"chisel2015-table-ii","chisel2015:Table II","Table II",[315,317,318,320],{"label":316,"unit":119,"statistic":120,"alignment":121},"Meshing Time (ms.)",{"label":316,"unit":119,"statistic":123,"alignment":121},{"label":319,"unit":119,"statistic":120,"alignment":121},"Update Time (ms.)",{"label":319,"unit":119,"statistic":123,"alignment":121},[322],{"dataset":126,"sequence":127,"environment":128},[324,326],{"name":325,"methodId":67,"linkable":73,"proposed":73,"self":73},"256^3 Fixed Grid (baseline)",{"name":327,"methodId":5,"linkable":77,"proposed":77,"self":77},"16^3 Spatial Hashing (CHISEL)",[329,331,333,335,337,339,341,342],[148,148,148,330,150,148,148,150,148],2067,[148,152,148,332,150,148,148,150,148],679,[148,153,148,334,150,148,148,150,148],3769,[148,175,148,336,150,148,148,150,148],1279,[152,148,148,338,150,148,148,150,148],102,[152,152,148,340,150,148,148,150,148],25,[152,153,148,208,150,148,148,150,148],[152,175,148,210,150,148,148,150,148],[],[313],[346],"not stated",[],[349],"Per-frame mesh generation and TSDF update (colorization, space carving, projection mapping) on the 'Room' dataset; platform not stated (update time equals the tablet value in Fig. 9e)",[351,357],{"group":352,"slug":353,"sourceLabel":6,"table":354,"selfRows":175,"datasets":355},"chisel2015:Text Sec.IV-D","chisel2015-text-sec-iv-d","Text Sec.IV-D",[98,356],"authors' Dragon dataset",{"group":358,"slug":359,"sourceLabel":6,"table":360,"selfRows":152,"datasets":361},"chisel2015:Text Sec.III-J","chisel2015-text-sec-iii-j","Text Sec.III-J",[362],"authors' office corridor",1790510657596]