[{"data":1,"prerenderedAt":468},["ShallowReactive",2],{"method-magnusson2007ndt3d":3},{"method":4,"reference":52,"equipment":72,"figures":107,"results":108},{"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":21,"limitations":26,"sensors":32,"platform":37,"estimator":40,"association":41,"timeModel":42,"deskew":43,"loopClosure":44,"globalOptimization":44,"mapRepresentation":45,"prior":46,"outputGeometry":47,"compute":48,"codeUrl":49,"codeLicense":50,"relatedVersions":51},"magnusson2007ndt3d","Magnusson et al., 2007","3D-NDT","Scan registration for autonomous mining vehicles using 3D-NDT",2007,"classic","C02","registration_component","作者把 Biber 與 Strasser 的二維 NDT 推廣為三維：把模型掃描切成固定格網，每格以點的平均與共變異數表示常態分布，再以牛頓法最佳化資料點落在分布上的分數，不需最近鄰搜尋。論文比較取樣方式、格子大小，以及八叉樹、加成式、迭代式細分、連結格與無限外界等變體，並以 Kvarntorp 礦坑的原型雷射與 SICK LMS 200 資料對照 ICP：迭代細分加無限外界的 3D-NDT 在 50 對機器人掃描中高精度配準 45 對；在相同取樣比例下，3D-NDT 通常比 ICP 快將近三倍，但起始誤差較大時比 ICP 更早失敗。","3D generalization of NDT (cell-wise normal distributions scored with Newton's method, no nearest-neighbour search) with sampling, cell-size and subdivision variants, compared with ICP on Kvarntorp mine data: iterative-subdivision NDT registered 45 of 50 robot scan pairs with high accuracy, 3D-NDT was typically almost three times faster than ICP at the same sampling ratio, but began failing at slightly smaller initial errors.","full_text_reviewed","peer_reviewed_published","main_body","論文的目標應用是在隧道開挖用的鑽堡上量測隧道斷面，用來檢核新隧道形狀、估算開挖量並檢查舊隧道安全（Sec. 1）；實驗地點為瑞典 Kvarntorp 已停產的砂岩礦坑（Sec. 5.1）。屬地下工程與隧道施工相關場景，但不是建築工地。JUNCTION 兩站在同一位置掃描，參考位姿即為零位移；TUNNEL 與 KVARNTORP-LOOP 的參考位姿是多次配準後目視挑選最佳結果再取平均，並非獨立量測，全測站量測精度不足，只用作初始估計。",[20],"underground_or_tunnel",[22,23,24,25],"faster and slightly more reliable than ICP on the mine data (Abstract)","iterative subdivision with infinite outer bounds registered 45 of 50 KVARNTORP-LOOP pairs with high accuracy and failed on two (Sec. 5.2.2, Fig. 21)","lower median error than ICP in most single-pair tests (Sec. 5.2.1)","NDT storage needs only a small fraction of the space of point clouds (Sec. 6)",[27,28,29,30,31],"cell size must be chosen: too small cells fail from larger initial errors and too large cells blur features; about 1 to 2 m suited this mine (Sec. 4.2, 5.2.1)","failures start at smaller initial errors than ICP on JUNCTION (3D-NDT from 2 m or 0.3 rad, ICP reliable to 2.5 m or 0.35 rad) (Sec. 5.2.1)","some non-converging outlier runs; at low sample ratios 3D-NDT failed up to 12% sampling while ICP was acceptable down to about 8% (Sec. 5.2.1)","reference poses for TUNNEL and KVARNTORP-LOOP were chosen by inspecting registration attempts; no independent ground truth (Sec. 5.1, 5.2)","array storage of all cells caused large memory use and swapping with 0.5 m cells (Sec. 5.2.1)",[33,34,35,36],"Optab Optronikinnovation AB prototype 3D laser range finder (modulated infrared laser on a rotating mirror, phase-shift ranging; pitching scans for TUNNEL, yawing scans for JUNCTION)","SICK LMS 200 2D laser scanner on a pan-tilt unit giving pitching 3D scans of about 95,000 points (KVARNTORP-LOOP)","2D wheel odometry of the robot for initial pose estimates (KVARNTORP-LOOP)","total station measuring three marked points on the scanner (TUNNEL); not accurate enough as ground truth, used only as initial estimate",[38,39],"static Optab prototype scanner (JUNCTION: two scans from the same pose at different resolutions; TUNNEL: two poses about 4 m apart); carrier not described","Tjorven mobile robot, driven manually and stationary during each scan (KVARNTORP-LOOP)","Newton's method with line search on the negated NDT score (maximum step 0.05, convergence when the change of p is below 0.0001); 7-parameter axis-angle transform (Eq. 13), noted as redundant; ICP baseline: point-to-point least squares with a 1 m fixed outlier threshold and approximate kd-tree search (ANN)","each data-scan point is scored against the normal distribution of the cell it falls in (point-to-distribution); variants: octree, additive and iterative subdivision, linked cells and infinite outer bounds","not_applicable (pairwise scan registration)","not_applicable (robot kept stationary during each scan)","none","regular grid of cells (1 m baseline, 0.5 m to 3 m tested) storing mean and covariance of the points in each cell that holds more than a minimum number of points (Sec. 2.2 calls five points per cell a reasonable limit); octree-forest, additive and iterative subdivision variants","initial pose estimate required: synthetic offsets around a reference pose (JUNCTION, TUNNEL), total-station estimate (TUNNEL), 2D odometry (KVARNTORP-LOOP)","6-DoF rigid transformation","C++ implementation (ANN library for ICP, newmat for NDT) on an AMD Athlon at 1950 MHz with 512 MB memory; ICP about three times slower than 3D-NDT in the sample-ratio tests and on KVARNTORP-LOOP; iterative 3D-NDT about twice a single-resolution run",null,"not_verified",[],{"id":5,"kind":53,"shortName":7,"title":8,"authors":54,"year":9,"venue":58,"venueType":59,"publisher":60,"volumeIssuePages":61,"doi":62,"arxivId":49,"url":63,"firstPublicDate":64,"publicationStatus":16,"metadataStatus":65,"fulltextStatus":15,"era":10,"classicReason":66,"codeUrl":49,"cluster":11,"topics":67,"mdpi":68,"verification":69,"label":6,"fulltextRoute":70,"versionRead":71,"addedByCensus":68},"method",[55,56,57],"Martin Magnusson","Achim Lilienthal","Tom Duckett","Journal of Field Robotics","journal","Wiley","24(10):803-827","10.1002\u002Frob.20204","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1002\u002Frob.20204","2007-10-24","metadata_verified","principle reused and underground relevance: generalizes Biber-Strasser NDT to 3D and evaluates it against 3D ICP on mine data.",[11],false,"confirmed","NTU institutional (Chrome)","version of record PDF, Journal of Field Robotics 24(10):803-827 (Wiley), downloaded through NTU institutional access",[73,80,85,90,95,101],{"category":74,"model":75,"canonical":75,"role":76,"dataset":77,"specs":78,"locator":79},"lidar","Optab prototype 3D laser range finder","method input","JUNCTION and TUNNEL","modulated infrared laser projected onto a rotating mirror, phase-shift ranging; JUNCTION data scan 139,642 and model 72,417 points; TUNNEL about 27,500 points per scan","Sec. 5.1; Fig. 9",{"category":74,"model":81,"canonical":81,"role":76,"dataset":82,"specs":83,"locator":84},"SICK LMS 200","KVARNTORP-LOOP","2D scanner on a pan-tilt unit for pitching 3D scans; about 95,000 points per scan; scans 4 to 5 m apart","Sec. 5.1; Fig. 10",{"category":86,"model":87,"canonical":87,"role":76,"dataset":82,"specs":88,"locator":89},"platform","Tjorven (authors' mobile robot)","driven manually, stationary during scans; also carries a digital camera, sonar array, omnidirectional camera and differential GPS, not used for registration","Sec. 5.1; Fig. 10 caption",{"category":91,"model":92,"canonical":92,"role":76,"dataset":82,"specs":93,"locator":94},"wheel_or_leg_odometry","Tjorven 2D odometry","pose error up to about 1.5 m and 0.2 rad between scans; reset after scans 11, 16 and 66","Sec. 5.2.2",{"category":96,"model":97,"canonical":97,"role":76,"dataset":98,"specs":99,"locator":100},"total_station","total station (model not stated)","TUNNEL","tripod-mounted; measured three marked points on the scanner from a fixed position; not accurate enough for ground truth, used as initial estimate","Sec. 5.1; Fig. 6",{"category":102,"model":103,"canonical":103,"role":104,"dataset":49,"specs":105,"locator":106},"compute","AMD Athlon 1950 MHz with 512 MB memory","compute for runtime","runs all timing experiments","Sec. 5 baseline",[],{"totalRows":109,"groupCount":110,"groups":111,"others":407},55,14,[112,223,281,357],{"slug":113,"group":114,"sourceId":115,"sourceLabel":116,"table":117,"selfRows":118,"metrics":119,"seqs":128,"entrants":149,"cells":169,"outcomes":216,"locators":217,"hardware":218,"wordings":220,"notes":221},"pang2018ndticp-table-iii","pang2018ndticp:Table III","pang2018ndticp","Pang et al., 2018","Table III",18,[120,124],{"label":121,"unit":122,"statistic":123,"alignment":44},"Localization MAE Error (m)","m","mean",{"label":125,"unit":126,"statistic":123,"alignment":127},"Average time for registration (ms)","ms","not_applicable",[129,133,135,137,139,141,143,145,147],{"dataset":130,"sequence":131,"environment":132},"MCity test route (350 m)","voxel size (m) = 0.5","MCity proving ground, Ann Arbor (synthetic urban and suburban test site, no moving objects during tests)",{"dataset":130,"sequence":134,"environment":132},"voxel size (m) = 1",{"dataset":130,"sequence":136,"environment":132},"voxel size (m) = 2",{"dataset":130,"sequence":138,"environment":132},"transformation difference threshold (m) = 0.005",{"dataset":130,"sequence":140,"environment":132},"transformation difference threshold (m) = 0.01",{"dataset":130,"sequence":142,"environment":132},"transformation difference threshold (m) = 0.02",{"dataset":130,"sequence":144,"environment":132},"maximum step size (m) = 0.05",{"dataset":130,"sequence":146,"environment":132},"maximum step size (m) = 0.1",{"dataset":130,"sequence":148,"environment":132},"maximum step size (m) = 0.2",[150,153,155,157,159,161,163,165,167],{"name":151,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT (voxel size (m) = 0.5)",true,{"name":154,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT (voxel size (m) = 1)",{"name":156,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT (voxel size (m) = 2)",{"name":158,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT (transformation difference threshold (m) = 0.005)",{"name":160,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT (transformation difference threshold (m) = 0.01)",{"name":162,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT (transformation difference threshold (m) = 0.02)",{"name":164,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT (maximum step size (m) = 0.05)",{"name":166,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT (maximum step size (m) = 0.1)",{"name":168,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT (maximum step size (m) = 0.2)",[170,174,177,179,181,184,186,189,191,194,196,199,201,204,206,209,211,214],[171,171,171,172,173,171,173,173,171],0,0.0525,-1,[171,175,171,176,173,171,171,173,171],1,11.799,[175,171,175,178,173,171,173,173,171],0.0463,[175,175,175,180,173,171,171,173,171],11.088,[182,171,182,183,173,171,173,173,171],2,0.0501,[182,175,182,185,173,171,171,173,171],9.78,[187,171,187,188,173,171,173,173,171],3,0.0473,[187,175,187,190,173,171,171,173,171],13.16,[192,171,192,193,173,171,173,173,171],4,0.0452,[192,175,192,195,173,171,171,173,171],11.09,[197,171,197,198,173,171,173,173,171],5,0.0532,[197,175,197,200,173,171,171,173,171],9.986,[202,171,202,203,173,171,173,173,171],6,0.0472,[202,175,202,205,173,171,171,173,171],12.937,[207,171,207,208,173,171,173,173,171],7,0.0467,[207,175,207,210,173,171,171,173,171],11.082,[212,171,212,213,173,171,173,173,171],8,0.0581,[212,175,212,215,173,171,171,173,171],9.874,[],[117],[219],"not stated for the timings; the vehicle computer has a Core i7 CPU with Ubuntu and ROS (Sec. IV-A)",[],[222],"MCity 350 m route; NDT with one parameter varied per row (values of the other parameters not stated)",{"slug":224,"group":225,"sourceId":115,"sourceLabel":116,"table":226,"selfRows":212,"metrics":227,"seqs":231,"entrants":240,"cells":246,"outcomes":272,"locators":274,"hardware":277,"wordings":278,"notes":279},"pang2018ndticp-table-i","pang2018ndticp:Table I","Table I",[228,230],{"label":229,"unit":122,"statistic":123,"alignment":44},"Localization MAE error (m)",{"label":125,"unit":126,"statistic":123,"alignment":127},[232,234,236,238],{"dataset":130,"sequence":233,"environment":132},"reference map resolution 9 points\u002Fm2",{"dataset":130,"sequence":235,"environment":132},"reference map resolution 36 points\u002Fm2",{"dataset":130,"sequence":237,"environment":132},"reference map resolution 121 points\u002Fm2",{"dataset":130,"sequence":239,"environment":132},"reference map resolution 400 points\u002Fm2",[241,243],{"name":242,"methodId":5,"linkable":152,"proposed":68,"self":152},"NDT",{"name":244,"methodId":245,"linkable":152,"proposed":68,"self":68},"ICP (kd-tree, point-to-point)","besl1992icp",[247,249,251,253,255,257,259,261,263,264,266,267,269,271],[171,171,171,248,173,171,173,173,171],0.3391,[171,175,171,250,173,171,171,173,171],5.814,[175,171,171,252,173,175,173,173,171],0.125,[175,175,171,254,173,175,171,173,171],68.726,[171,171,175,256,173,171,173,173,171],0.0618,[171,175,175,258,173,171,171,173,171],9.648,[175,171,175,260,173,175,173,173,171],0.1167,[175,175,175,262,173,175,171,173,171],203.549,[171,171,182,188,173,171,173,173,171],[171,175,182,265,173,171,171,173,171],10.404,[175,171,182,49,171,182,173,173,171],[171,171,187,268,173,171,173,173,171],0.047,[171,175,187,270,173,171,171,173,171],11.824,[175,171,187,49,171,182,173,173,171],[273],"not reported: ICP rows for 121 points\u002Fm2 or higher are omitted from Table I because each registration step takes more than 2 s",[226,275,276],"Table I and note","Table I footnote",[219],[],[280],"MCity route of 350 m at 17 mph; localisation MAE and average registration time for different reference-map resolutions; ICP rows for 121 and 400 points\u002Fm2 not listed because each registration took more than 2 s",{"slug":282,"group":283,"sourceId":5,"sourceLabel":6,"table":284,"selfRows":202,"metrics":285,"seqs":304,"entrants":320,"cells":329,"outcomes":339,"locators":343,"hardware":349,"wordings":351,"notes":352},"magnusson2007ndt3d-text-sec-5-2-1","magnusson2007ndt3d:Text Sec. 5.2.1","Text Sec. 5.2.1",[286,289,292,295,298,301],{"label":287,"unit":122,"statistic":288,"alignment":44},"initial translation error at which failed registrations start","not_reported",{"label":290,"unit":291,"statistic":288,"alignment":44},"initial rotation error at which failed registrations start","rad",{"label":293,"unit":294,"statistic":288,"alignment":44},"failed registrations occurred up to this sample ratio","% of data-scan points",{"label":296,"unit":297,"statistic":288,"alignment":127},"ICP execution time around three times longer than 3D-NDT","x (time ratio)",{"label":299,"unit":300,"statistic":288,"alignment":44},"initial poses from which additive octree subdivision failed","count of 100 poses",{"label":302,"unit":303,"statistic":288,"alignment":44},"accurately registered from at least this share of initial poses (only iterative subdivision and additive subdivision with infinite outer bounds)","%",[305,309,311,313,315,317],{"dataset":306,"sequence":307,"environment":308},"JUNCTION","initial translation error test (rotation error 0)","underground mine tunnel end section (Kvarntorp)",{"dataset":306,"sequence":310,"environment":308},"initial rotation error test (translation error 0)",{"dataset":306,"sequence":312,"environment":308},"sample ratio tests (et = 1 m, er = 0.1 rad)",{"dataset":77,"sequence":314,"environment":308},"sample ratio tests",{"dataset":306,"sequence":316,"environment":308},"discretization tests (100 initial poses)",{"dataset":98,"sequence":318,"environment":319},"discretization tests","underground mine tunnel (Kvarntorp), scans about 4 m apart",[321,323,325,327],{"name":322,"methodId":5,"linkable":152,"proposed":152,"self":152},"3D-NDT (baseline, fixed 1 m cells)",{"name":324,"methodId":5,"linkable":152,"proposed":152,"self":152},"ICP vs 3D-NDT",{"name":326,"methodId":5,"linkable":152,"proposed":152,"self":152},"3D-NDT, additive octree subdivision",{"name":328,"methodId":5,"linkable":152,"proposed":152,"self":152},"3D-NDT, iterative subdivision; additive subdivision with infinite outer bounds",[330,331,333,335,336,337],[171,171,171,182,171,171,173,173,171],[171,175,175,332,171,175,173,173,171],0.3,[171,182,182,334,173,182,173,173,171],12,[175,187,187,187,175,187,171,173,175],[182,192,192,182,173,192,173,173,182],[187,197,197,338,182,192,173,173,187],75,[340,341,342],"failures begin","approximate","lower bound (at least 75%)",[344,345,346,347,348],"Sec. 5.2.1; Fig. 16","Sec. 5.2.1; Fig. 17","Sec. 5.2.1; Fig. 12","Sec. 5.2.1; Figs. 12-13","Sec. 5.2.1; Fig. 18",[350],"AMD Athlon 1950 MHz, 512 MB RAM",[],[353,354,355,356],"JUNCTION pair (Optab prototype scanner, both scans from the same pose, ground truth = zero motion); 100 runs per setting from start poses on a sphere around the reference; baseline 10% spatially distributed sampling of the data scan","sample-ratio tests on the JUNCTION and TUNNEL pairs (Figs. 12-13); timing statement from the text; baseline settings otherwise","discretization tests; JUNCTION with et = 1 m and er = 0.2 rad","discretization tests; TUNNEL with et = 1 m and er = 0.1 rad; reference pose chosen visually",{"slug":358,"group":359,"sourceId":5,"sourceLabel":6,"table":360,"selfRows":202,"metrics":361,"seqs":376,"entrants":380,"cells":389,"outcomes":398,"locators":400,"hardware":403,"wordings":404,"notes":405},"magnusson2007ndt3d-text-sec-5-2-2","magnusson2007ndt3d:Text Sec. 5.2.2","Text Sec. 5.2.2",[362,365,367,369,371,374],{"label":363,"unit":364,"statistic":288,"alignment":44},"successful registrations (octree subdivision, 2 m cells split to 1 and 0.5 m)","count of 50 pairs",{"label":366,"unit":364,"statistic":288,"alignment":44},"registered with very high accuracy (iterative subdivision)",{"label":368,"unit":364,"statistic":288,"alignment":44},"'acceptable' matches (iterative subdivision)",{"label":370,"unit":364,"statistic":288,"alignment":44},"failed registrations (scan 49 position, scan 41 orientation)",{"label":372,"unit":373,"statistic":288,"alignment":127},"3D-NDT typically almost three times faster than ICP at the same sampling ratio","x (speed ratio)",{"label":375,"unit":297,"statistic":288,"alignment":127},"iterative 3D-NDT took about twice as long as a single iteration of the other versions",[377],{"dataset":82,"sequence":378,"environment":379},"scans 17-66","underground mine tunnels (Kvarntorp)",[381,383,385,387],{"name":382,"methodId":5,"linkable":152,"proposed":152,"self":152},"3D-NDT, octree subdivision",{"name":384,"methodId":5,"linkable":152,"proposed":152,"self":152},"3D-NDT, iterative subdivision with infinite outer bounds",{"name":386,"methodId":5,"linkable":152,"proposed":152,"self":152},"3D-NDT vs ICP",{"name":388,"methodId":5,"linkable":152,"proposed":152,"self":152},"3D-NDT iterative vs single-resolution variants",[390,392,394,395,396,397],[171,171,171,391,173,171,173,173,171],40,[175,175,171,393,173,175,173,173,171],45,[175,182,171,187,173,182,173,173,171],[175,187,171,182,173,175,173,173,171],[182,192,171,187,171,171,171,173,171],[187,197,171,182,173,171,171,173,171],[399],"approximate ('almost three times')",[94,401,402],"Sec. 5.2.2; Fig. 21 caption","Fig. 21 caption",[350],[],[406],"KVARNTORP-LOOP scans 17-66 (50 consecutive pairs), SICK LMS 200 on Tjorven; 8000 random data-scan samples (about 8%), all model points; initial poses from 2D odometry; 'good' = within 0.10 m and 0.005 rad, 'acceptable' = within 0.20 m and 0.010 rad of reference poses obtained by inspecting a number of registration attempts and averaging the best matches",[408,414,420,427,434,440,445,451,456,464],{"group":409,"slug":410,"sourceLabel":116,"table":411,"selfRows":187,"datasets":412},"pang2018ndticp:Text Sec. IV-E","pang2018ndticp-text-sec-iv-e","Text Sec. IV-E",[413],"MCity route (1 km)",{"group":415,"slug":416,"sourceLabel":116,"table":417,"selfRows":187,"datasets":418},"pang2018ndticp:Text Sec. IV-F","pang2018ndticp-text-sec-iv-f","Text Sec. IV-F",[419],"MCity",{"group":421,"slug":422,"sourceLabel":423,"table":424,"selfRows":182,"datasets":425},"magnusson2009icpndt:Text Fig. 7 caption","magnusson2009icpndt-text-fig-7-caption","Magnusson et al., 2009","Text Fig. 7 caption",[426],"Kvarntorp data set A",{"group":428,"slug":429,"sourceLabel":430,"table":431,"selfRows":182,"datasets":432},"magnusson2015beyondpoints:Fig. 3 (execution-time table)","magnusson2015beyondpoints-fig-3-execution-time-table","Magnusson et al., 2015","Fig. 3 (execution-time table)",[433],"ETH Challenging Laser Registration (six data sets)",{"group":435,"slug":436,"sourceLabel":116,"table":437,"selfRows":182,"datasets":438},"pang2018ndticp:Table IV","pang2018ndticp-table-iv","Table IV",[439],"MSU West Circle Drive",{"group":441,"slug":442,"sourceLabel":423,"table":443,"selfRows":175,"datasets":444},"magnusson2009icpndt:Text Fig. 8 caption","magnusson2009icpndt-text-fig-8-caption","Text Fig. 8 caption",[426],{"group":446,"slug":447,"sourceLabel":423,"table":448,"selfRows":175,"datasets":449},"magnusson2009icpndt:Text Sec. IV-D-2","magnusson2009icpndt-text-sec-iv-d-2","Text Sec. IV-D-2",[450],"Kvarntorp data set B",{"group":452,"slug":453,"sourceLabel":423,"table":454,"selfRows":175,"datasets":455},"magnusson2009icpndt:Text Sec. IV-D-3","magnusson2009icpndt-text-sec-iv-d-3","Text Sec. IV-D-3",[450],{"group":457,"slug":458,"sourceLabel":430,"table":459,"selfRows":175,"datasets":460},"magnusson2015beyondpoints:Text Sec. V-A","magnusson2015beyondpoints-text-sec-v-a","Text Sec. V-A",[461,462,463],"Gazebo (winter)","Stairs","Wood (summer)",{"group":465,"slug":466,"sourceLabel":116,"table":448,"selfRows":175,"datasets":467},"pang2018ndticp:Text Sec. IV-D-2","pang2018ndticp-text-sec-iv-d-2",[439],1790510657190]