[{"data":1,"prerenderedAt":514},["ShallowReactive",2],{"method-stoyanov2012d2dndt":3},{"method":4,"reference":55,"equipment":76,"figures":90,"results":91},{"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":29,"sensors":39,"platform":41,"estimator":42,"association":43,"timeModel":44,"deskew":45,"loopClosure":46,"globalOptimization":47,"mapRepresentation":48,"prior":49,"outputGeometry":50,"compute":51,"codeUrl":52,"codeLicense":53,"relatedVersions":54},"stoyanov2012d2dndt","Stoyanov et al., 2012","D2D-NDT","Fast and accurate scan registration through minimization of the distance between compact 3D NDT representations",2012,"classic","C02","registration_component","此研究把固定與移動兩片掃描都轉成三維常態分布轉換（3D-NDT）模型，也就是在規則網格的每個格子以一個高斯分布描述局部表面，再直接最小化兩個模型之間的 L2 距離（分布對分布，D2D），不像點對分布（P2D）或 ICP 那樣逐點計算。目標函數只在兩模型彼此最近的高斯成分之間評估，具有解析梯度與 Hessian，以牛頓法搭配 More-Thuente 線搜尋求解，並在 4、2、1、0.5 m 的網格上逐層配準；每次迭代直接轉換整個 NDT 模型並累積齊次轉換矩陣，以處理 SE(3) 的結構。作者另以 3D-NDT 直方圖對齊主要平面法向量來估計初始旋轉，並依 Censi 的方法推導封閉形式共變異數。在 AASS 室內迴圈與 Hannover2 戶外資料上，D2D 的精度與 ICP、P2D 相當，在 AASS 上快將近一個數量級；模擬中直方圖初始化平均約 150 ms，FPFH 約 15 秒。結果多以箱形圖呈現且未報告硬體；在稀疏的戶外資料上各方法都有大量失敗，共變異數估計也偏大。","D2D-NDT converts both scans into 3D-NDT Gaussian grids and minimizes the L2 distance between the two mixtures over pairwise closest components with Newton's method and analytic derivatives on a 4, 2, 1, 0.5 m grid cascade; a 3D-NDT histogram supplies an initial rotation and a Censi-style closed-form covariance is derived; on AASS and Hannover2 it matches ICP and P2D accuracy while running up to about an order of magnitude faster.","full_text_reviewed","peer_reviewed_published","supplementary","not_reported（測試資料為 AASS 室內迴圈、Hannover2 戶外掃描，以及模擬的辦公室與瀝青廠場景，論文未討論營建應用）",[20,21],"public_benchmark","simulation",[23,24,25,26,27,28],"D2D final alignments comparable to ICP and NDT-P2D on the AASS loop, 59 scan pairs (Sec. 6.2, Fig. 5)","with histogram initialization D2D has the largest share of successfully registered AASS pairs, success being at most 0.5 m and 0.2 rad error (Sec. 6.2, Fig. 7)","D2D consistently faster than ICP and P2D (Sec. 6.2)","in the 343-offset robustness test on five AASS pairs (success within 0.2 m and 0.05 rad) D2D slightly outperforms P2D with stable results, while ICP fails on corridor pairs 1 and 5 (Sec. 6.2, Fig. 9)","histogram initialization more accurate and much faster than FPFH (Sec. 6.1, Fig. 4)","inexpensive closed-form covariance (Sec. 5, 7)",[30,31,32,33,34,35,36,37,38],"as an iterative method it remains susceptible to local minima (Sec. 7)","without initialization all methods show large translational error variance on AASS, with failures in corridors and locally similar geometry (Sec. 6.2)","on the sparser outdoor Hannover2 set all six combinations are worse, with many outliers and an apparent systematic bias, and histogram initialization does not help and slightly lowers P2D success (Sec. 6.2, Figs. 6, 8)","the histogram initial guess corrects orientation only (Sec. 6.1)","the covariance estimate over-estimates the observed covariance (Sec. 6.3)","no comparison with Generalized ICP, 3DTK ICP or global registration methods (Sec. 6.2)","evaluation on scan pairs, not full trajectories, with results only as box plots (Sec. 6.2)","not tested on noisier or narrower field-of-view sensors (Sec. 7)","NDT variants depend on voxel-size choice (koide2021vgicp Sec. II, secondary)",[40],"[\"rotating SICK laser (AASS loop and Hannover2 data sets)\", \"simulated 3D range sensor in ROS\u002FGazebo with SICK LMS 200 error models, 180 x 120 deg field of view\"]",[],"minimizes an L2 distance between the fixed and moving 3D-NDT models, written as a sum of negative Gaussian terms over component pairs (Eq. 18); analytic gradient and Hessian; Newton's method with More-Thuente line search; parameters d1 = 1, d2 = 0.05; optimization repeated at grid sizes of 4, 2, 1 and 0.5 m; each pose increment is applied by transforming the whole moving NDT model and increments are accumulated in a homogeneous matrix so derivatives are always taken at the zero pose","distribution-to-distribution: each Gaussian component of the moving 3D-NDT is paired only with the closest component of the fixed 3D-NDT; no point-level correspondences; baselines ICP (PCL) and NDT-P2D run on point sets sub-sampled on a 0.1 m grid","not_applicable","not_reported","none","none; pairwise scan registration only; coupling with loop detection and a global pose-graph optimization is stated as future work","3D-NDT: one Gaussian (mean and covariance) per occupied cell of a regular grid, built for both scans at cell sizes 4, 2, 1 and 0.5 m; about 1,500 components at 0.5 m for a typical AASS scan","optional initial rotation from the 3D-NDT histogram: per range band 1 linear, 20 flat (orientation) and 5 spherical bins, three range bands, 78 values; the n = 3 dominant flat-patch directions of the two histograms are matched over 6 permutations and the rotation with the most similar histogram is kept; orientation only, no translation; in registration the top three candidate rotations and the zero pose each start a run and the best score is kept (up to four times the runtime)","6-DoF rigid transformation (homogeneous matrix) plus a closed-form covariance derived after Censi (2007) with the NDT component covariances as the measurements; in the simulation test the estimate over-estimated the sample covariance, which the authors consider acceptable","hardware not reported; NDT-histogram initialization averages on the order of 150 ms vs about 15 s for the ROS FPFH baseline (Sec. 6.1); D2D runs almost an order of magnitude faster than PCL ICP and the point-to-distribution variant on AASS (the sentence names 'D2D' for the second method, evidently a typo for P2D) and is consistently faster on Hannover as well (Figs. 5c, 6c, text only); objective evaluation costs O(q log q) for q Gaussian components vs O(n log n) for ICP, and D2D time is usually dominated by building the NDT models",null,"not_verified",[],{"id":5,"kind":56,"shortName":7,"title":8,"authors":57,"year":9,"venue":62,"venueType":63,"publisher":64,"volumeIssuePages":65,"doi":66,"arxivId":52,"url":67,"firstPublicDate":68,"publicationStatus":16,"metadataStatus":69,"fulltextStatus":15,"era":10,"classicReason":70,"codeUrl":52,"cluster":11,"topics":71,"mdpi":72,"verification":73,"label":6,"fulltextRoute":74,"versionRead":75,"addedByCensus":72},"method",[58,59,60,61],"Todor Stoyanov","Martin Magnusson","Henrik Andreasson","Achim J. Lilienthal","The International Journal of Robotics Research","journal","SAGE","31(12):1377-1393","10.1177\u002F0278364912460895","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1177\u002F0278364912460895","2012-09-24","metadata_verified","necessary technical node: distribution-to-distribution NDT that koide2021vgicp explicitly contrasts with (Sec. II).",[11],false,"confirmed","NTU institutional (Chrome)","SAGE version of record PDF, IJRR 31(12):1377-1393 (17 pages, 'Available access' via National Taiwan University Library)",[77,84],{"category":78,"model":79,"canonical":79,"role":80,"dataset":81,"specs":82,"locator":83},"lidar","rotating SICK laser","dataset sensor","AASS loop and Hannover2 (3D scans online repository)","real-world data sets acquired with SICK laser scanners on a rotating mount; AASS: 60 point clouds of about 90,000 points; Hannover: 923 point clouds of about 15,000 points (Table 1)","Sec. 6.1, 6.2, Table 1",{"category":78,"model":85,"canonical":85,"role":86,"dataset":87,"specs":88,"locator":89},"SICK LMS 200","method input","simulated Willow garage office world and asphalt-mill environment (ROS\u002FGazebo)","simulated only: the noise and mixed-measurement error models reported for this scanner (Ye and Borenstein 2002; Tuley et al. 2005) were implemented in ROS\u002FGazebo for a simulated 3D sensor with a 180 x 120 deg field of view","Sec. 6.1, 6.3",[],{"totalRows":92,"groupCount":93,"groups":94,"others":513},27,4,[95,379,441,487],{"slug":96,"group":97,"sourceId":98,"sourceLabel":99,"table":100,"selfRows":101,"metrics":102,"seqs":107,"entrants":154,"cells":168,"outcomes":373,"locators":374,"hardware":375,"wordings":376,"notes":377},"mrsmap2014-table-1","mrsmap2014:Table 1","mrsmap2014","Stückler & Behnke, 2014","Table 1",22,[103],{"label":104,"unit":105,"statistic":106,"alignment":45},"median relative pose error (RPE) in mm","mm","median",[108,112,114,116,118,120,122,124,126,128,130,132,134,136,138,140,142,144,146,148,150,152],{"dataset":109,"sequence":110,"environment":111},"TUM RGB-D (Freiburg)","fr1 360","indoor office and structure\u002Ftexture test scenes, RGB-D camera (carrying mode not stated in the paper)",{"dataset":109,"sequence":113,"environment":111},"fr1 desk",{"dataset":109,"sequence":115,"environment":111},"fr1 desk2",{"dataset":109,"sequence":117,"environment":111},"fr1 floor",{"dataset":109,"sequence":119,"environment":111},"fr1 plant",{"dataset":109,"sequence":121,"environment":111},"fr1 room",{"dataset":109,"sequence":123,"environment":111},"fr1 rpy",{"dataset":109,"sequence":125,"environment":111},"fr1 teddy",{"dataset":109,"sequence":127,"environment":111},"fr1 xyz",{"dataset":109,"sequence":129,"environment":111},"fr2 desk",{"dataset":109,"sequence":131,"environment":111},"fr2 large no loop",{"dataset":109,"sequence":133,"environment":111},"fr2 rpy",{"dataset":109,"sequence":135,"environment":111},"fr2 xyz",{"dataset":109,"sequence":137,"environment":111},"fr3 long office household",{"dataset":109,"sequence":139,"environment":111},"fr3 nostruct. notext. far",{"dataset":109,"sequence":141,"environment":111},"fr3 nostruct. notext. near",{"dataset":109,"sequence":143,"environment":111},"fr3 nostruct. text. far",{"dataset":109,"sequence":145,"environment":111},"fr3 nostruct. text. near",{"dataset":109,"sequence":147,"environment":111},"fr3 struct. notext. far",{"dataset":109,"sequence":149,"environment":111},"fr3 struct. notext. near",{"dataset":109,"sequence":151,"environment":111},"fr3 struct. text. far",{"dataset":109,"sequence":153,"environment":111},"fr3 struct. text. near",[155,158,160,163,165],{"name":156,"methodId":98,"linkable":157,"proposed":157,"self":72},"Ours (MRSMap)",true,{"name":159,"methodId":52,"linkable":72,"proposed":72,"self":72},"Warp [17] (OpenCV)",{"name":161,"methodId":162,"linkable":157,"proposed":72,"self":72},"GICP [5]","segal2009gicp",{"name":164,"methodId":5,"linkable":157,"proposed":72,"self":157},"3D-NDT [7]",{"name":166,"methodId":167,"linkable":157,"proposed":72,"self":72},"Fovis [12]","fovis2017",[169,173,176,179,182,184,186,188,190,192,194,196,198,200,202,204,206,208,210,211,213,215,217,219,221,223,224,225,226,228,230,232,233,234,236,237,239,240,242,244,245,247,249,251,253,254,256,257,259,261,263,266,268,269,271,273,275,277,279,280,281,284,285,286,287,289,291,293,294,295,297,300,302,304,306,308,311,313,315,317,319,322,324,326,328,330,333,334,336,338,340,343,344,345,347,349,351,352,353,355,357,360,362,363,364,365,367,368,370,371],[170,170,170,171,172,170,172,172,170],0,5.1,-1,[174,170,170,175,172,170,172,172,170],1,5.9,[177,170,170,178,172,170,172,172,170],2,18.8,[180,170,170,181,172,170,172,172,170],3,7.8,[93,170,170,183,172,170,172,172,170],7.1,[170,170,174,185,172,170,172,172,170],4.4,[174,170,174,187,172,170,172,172,170],5.8,[177,170,174,189,172,170,172,172,170],10.2,[180,170,174,191,172,170,172,172,170],7.9,[93,170,174,193,172,170,172,172,170],6.3,[170,170,177,195,172,170,172,172,170],4.5,[174,170,177,197,172,170,172,172,170],6.2,[177,170,177,199,172,170,172,172,170],10.4,[180,170,177,201,172,170,172,172,170],8.2,[93,170,177,203,172,170,172,172,170],6.6,[170,170,180,205,172,170,172,172,170],4.9,[174,170,180,207,172,170,172,172,170],2.1,[177,170,180,209,172,170,172,172,170],5,[180,170,180,193,172,170,172,172,170],[93,170,180,212,172,170,172,172,170],2.6,[170,170,93,214,172,170,172,172,170],3.5,[174,170,93,216,172,170,172,172,170],4.2,[177,170,93,218,172,170,172,172,170],16.1,[180,170,93,220,172,170,172,172,170],7.4,[93,170,93,222,172,170,172,172,170],4.6,[170,170,209,214,172,170,172,172,170],[174,170,209,222,172,170,172,172,170],[177,170,209,189,172,170,172,172,170],[180,170,209,227,172,170,172,172,170],6.1,[93,170,209,229,172,170,172,172,170],5.4,[170,170,231,180,172,170,172,172,170],6,[174,170,231,171,172,170,172,172,170],[177,170,231,199,172,170,172,172,170],[180,170,231,235,172,170,172,172,170],6.8,[93,170,231,229,172,170,172,172,170],[170,170,238,216,172,170,172,172,170],7,[174,170,238,227,172,170,172,172,170],[177,170,238,241,172,170,172,172,170],21.3,[180,170,238,243,172,170,172,172,170],8.8,[93,170,238,183,172,170,172,172,170],[170,170,246,212,172,170,172,172,170],8,[174,170,246,248,172,170,172,172,170],4.1,[177,170,246,250,172,170,172,172,170],3.9,[180,170,246,252,172,170,172,172,170],5.2,[93,170,246,222,172,170,172,172,170],[170,170,255,207,172,170,172,172,170],9,[174,170,255,207,172,170,172,172,170],[177,170,255,258,172,170,172,172,170],6.7,[180,170,255,260,172,170,172,172,170],4.3,[93,170,255,262,172,170,172,172,170],2.5,[170,170,264,265,172,170,172,172,170],10,21.8,[174,170,264,267,172,170,172,172,170],20.5,[177,170,264,241,172,170,172,172,170],[180,170,264,270,172,170,172,172,170],32.1,[93,170,264,272,172,170,172,172,170],11,[170,170,272,274,172,170,172,172,170],1.6,[174,170,272,276,172,170,172,172,170],1.7,[177,170,272,278,172,170,172,172,170],1.3,[180,170,272,216,172,170,172,172,170],[93,170,272,276,172,170,172,172,170],[170,170,282,283,172,170,172,172,170],12,1.4,[174,170,282,177,172,170,172,172,170],[177,170,282,276,172,170,172,172,170],[180,170,282,93,172,170,172,172,170],[93,170,282,288,172,170,172,172,170],1.9,[170,170,290,212,172,170,172,172,170],13,[174,170,290,292,172,170,172,172,170],3.2,[177,170,290,181,172,170,172,172,170],[180,170,290,216,172,170,172,172,170],[93,170,290,296,172,170,172,172,170],3.7,[170,170,298,299,172,170,172,172,170],14,9.7,[174,170,298,301,172,170,172,172,170],40.4,[177,170,298,303,172,170,172,172,170],8.6,[180,170,298,305,172,170,172,172,170],13.8,[93,170,298,307,172,170,172,172,170],11.3,[170,170,309,310,172,170,172,172,170],15,15.2,[174,170,309,312,172,170,172,172,170],28.2,[177,170,309,314,172,170,172,172,170],12.5,[180,170,309,316,172,170,172,172,170],17.1,[93,170,309,318,172,170,172,172,170],11.2,[170,170,320,321,172,170,172,172,170],16,18.5,[174,170,320,323,172,170,172,172,170],19.2,[177,170,320,325,172,170,172,172,170],10.9,[180,170,320,327,172,170,172,172,170],18.6,[93,170,320,329,172,170,172,172,170],20.8,[170,170,331,332,172,170,172,172,170],17,11.5,[174,170,331,238,172,170,172,172,170],[177,170,331,335,172,170,172,172,170],8.9,[180,170,331,337,172,170,172,172,170],10.6,[93,170,331,339,172,170,172,172,170],7.3,[170,170,341,342,172,170,172,172,170],18,2.2,[174,170,341,303,172,170,172,172,170],[177,170,341,195,172,170,172,172,170],[180,170,341,346,172,170,172,172,170],2.9,[93,170,341,348,172,170,172,172,170],9.1,[170,170,350,207,172,170,172,172,170],19,[174,170,350,303,172,170,172,172,170],[177,170,350,346,172,170,172,172,170],[180,170,350,354,172,170,172,172,170],2.4,[93,170,350,356,172,170,172,172,170],9.3,[170,170,358,359,172,170,172,172,170],20,5.5,[174,170,358,361,172,170,172,172,170],8.1,[177,170,358,183,172,170,172,172,170],[180,170,358,229,172,170,172,172,170],[93,170,358,243,172,170,172,172,170],[170,170,366,292,172,170,172,172,170],21,[174,170,366,175,172,170,172,172,170],[177,170,366,369,172,170,172,172,170],5.6,[180,170,366,359,172,170,172,172,170],[93,170,366,372,172,170,172,172,170],6.5,[],[100],[],[],[378],"Incremental (frame-to-frame) registration on TUM Freiburg sequences; median translational relative pose error in mm (maximum values in brackets in the table not extracted); warp is the OpenCV reimplementation",{"slug":380,"group":381,"sourceId":382,"sourceLabel":383,"table":384,"selfRows":177,"metrics":385,"seqs":392,"entrants":397,"cells":410,"outcomes":431,"locators":432,"hardware":434,"wordings":438,"notes":439},"magnusson2015beyondpoints-fig-3-execution-time-table","magnusson2015beyondpoints:Fig. 3 (execution-time table)","magnusson2015beyondpoints","Magnusson et al., 2015","Fig. 3 (execution-time table)",[386,389],{"label":387,"unit":388,"statistic":106,"alignment":44},"execution time Q50","s",{"label":390,"unit":388,"statistic":391,"alignment":44},"execution time Q95","other: 95th percentile (Q95)",[393],{"dataset":394,"sequence":395,"environment":396},"ETH Challenging Laser Registration (six data sets)","all scan pairs and pose offsets","indoor and outdoor (apartment, stairs, hallway, park gazebo, forest, alpine plain)",[398,401,404,406,408],{"name":399,"methodId":400,"linkable":157,"proposed":72,"self":72},"Plane ICP (libpointmatcher point-to-plane baseline)","chen1992pointtoplane",{"name":402,"methodId":403,"linkable":157,"proposed":72,"self":72},"P2D-NDT (perception_oru)","magnusson2007ndt3d",{"name":405,"methodId":5,"linkable":157,"proposed":72,"self":157},"D2D-NDT (perception_oru)",{"name":407,"methodId":52,"linkable":72,"proposed":72,"self":72},"MUMC DC-OFF",{"name":409,"methodId":52,"linkable":72,"proposed":72,"self":72},"MUMC DC-ON",[411,413,415,417,419,421,423,425,427,429],[170,170,170,412,172,170,170,172,170],2.58,[170,174,170,414,172,170,170,172,170],8.43,[174,170,170,416,172,170,174,172,170],1.48,[174,174,170,418,172,170,174,172,170],8.75,[177,170,170,420,172,170,174,172,170],0.37,[177,174,170,422,172,170,174,172,170],0.82,[180,170,170,424,172,170,177,172,170],2.95,[180,174,170,426,172,170,177,172,170],18.01,[93,170,170,428,172,170,177,172,170],2.41,[93,174,170,430,172,170,177,172,170],12.95,[],[433],"Fig. 3 (table)",[435,436,437],"i7 2.2 GHz","i7 3.5 GHz","i7 3.4 GHz",[],[440],"execution time per registration over all six ETH data sets, including pre-processing and excluding file loading; single-threaded; different CPUs per method, so the authors call the comparison coarse",{"slug":442,"group":443,"sourceId":98,"sourceLabel":99,"table":444,"selfRows":177,"metrics":445,"seqs":450,"entrants":453,"cells":459,"outcomes":480,"locators":481,"hardware":482,"wordings":484,"notes":485},"mrsmap2014-table-2","mrsmap2014:Table 2","Table 2",[446],{"label":447,"unit":448,"statistic":449,"alignment":45},"average runtime in milliseconds","ms","mean",[451,452],{"dataset":109,"sequence":113,"environment":111},{"dataset":109,"sequence":129,"environment":111},[454,455,456,457,458],{"name":156,"methodId":98,"linkable":157,"proposed":157,"self":72},{"name":159,"methodId":52,"linkable":72,"proposed":72,"self":72},{"name":161,"methodId":162,"linkable":157,"proposed":72,"self":72},{"name":164,"methodId":5,"linkable":157,"proposed":72,"self":157},{"name":166,"methodId":167,"linkable":157,"proposed":72,"self":72},[460,462,464,466,468,470,472,474,476,478],[170,170,170,461,172,170,170,172,170],75.15,[174,170,170,463,172,170,170,172,170],108.64,[177,170,170,465,172,170,170,172,170],4015.4,[180,170,170,467,172,170,170,172,170],414.87,[93,170,170,469,172,170,170,172,170],15.98,[170,170,174,471,172,170,170,172,170],61.47,[174,170,174,473,172,170,170,172,170],99.29,[177,170,174,475,172,170,170,172,170],3147.3,[180,170,174,477,172,170,170,172,170],892.59,[93,170,174,479,172,170,170,172,170],12.98,[],[444],[483],"Intel Core i7 3610QM 2.3 GHz notebook CPU, VGA images",[],[486],"Average runtime per incremental registration in ms (standard deviations in the table not extracted)",{"slug":488,"group":489,"sourceId":5,"sourceLabel":6,"table":490,"selfRows":174,"metrics":491,"seqs":494,"entrants":499,"cells":502,"outcomes":505,"locators":507,"hardware":509,"wordings":510,"notes":511},"stoyanov2012d2dndt-text-sec-6-1","stoyanov2012d2dndt:Text Sec. 6.1","Text Sec. 6.1",[492],{"label":493,"unit":448,"statistic":449,"alignment":46},"average runtimes on the order of 150 ms",[495],{"dataset":496,"sequence":497,"environment":498},"simulated Willow and Terrain data sets (ROS\u002FGazebo)","160 scan pairs per data set","simulated indoor office and outdoor asphalt mill",[500],{"name":501,"methodId":5,"linkable":157,"proposed":157,"self":157},"3D-NDT histogram initialization",[503],[170,170,170,504,170,170,172,172,170],150,[506],"approximate, stated as on the order of 150 ms",[508],"Sec. 6.1, Fig. 4(b)",[],[],[512],"Initial orientation estimation on simulated scan pairs (20 positions per environment, 10 scans 15 deg apart, 8 pairs 30 deg apart per position, 160 pairs per data set, about 80% overlap); average runtime stated in text for both data sets; FPFH baseline is the ROS implementation",[],1790510664747]