[{"data":1,"prerenderedAt":146},["ShallowReactive",2],{"method-biber2003ndt":3},{"method":4,"reference":44,"equipment":63,"figures":76,"results":77},{"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":23,"sensors":27,"platform":29,"estimator":31,"association":32,"timeModel":33,"deskew":34,"loopClosure":35,"globalOptimization":36,"mapRepresentation":37,"prior":38,"outputGeometry":39,"compute":40,"codeUrl":41,"codeLicense":42,"relatedVersions":43},"biber2003ndt","Biber & Strasser, 2003","NDT (2D)","The normal distributions transform: a new approach to laser scan matching",2003,"classic","C02","registration_component","常態分布轉換（Normal Distributions Transform, NDT）將二維平面切成 100 cm 見方的網格，每個至少含三點的網格以點的平均與共變異數建立常態分布，並使用四組錯開半格的重疊網格降低離散化影響，使一次掃描成為分段連續且可微的機率密度。另一次掃描的點經轉換後在此密度上計分，再以牛頓法最佳化位姿，不需建立明確的點對應。作者以此進行相對於關鍵影格的位置追蹤，並以關鍵影格及其全域位姿構成地圖，利用成對匹配的 Hessian 建立二次誤差模型，只在新關鍵影格三條邊以內的子圖上最佳化。實驗以 SICK 雷射掃描儀在未改造的室內走廊、不使用里程計完成建圖，在 1.4 GHz 電腦上離線每秒約可處理 97 次掃描。","Represents a 2D scan as per-cell normal distributions forming a differentiable density; another scan is matched by Newton's method without explicit correspondences.","full_text_reviewed","peer_reviewed_published","main_body","not_reported（實驗僅在機器人所在的實驗室與走廊等室內環境進行，未涉及建物施工或工地）",[20],"completed_building",[22],"[\"no explicit correspondences between points or features are needed (abstract, Sec. IX)\", \"all derivatives are analytic, which is fast and correct (Sec. IX)\", \"real-time indoor mapping without odometry in the reported run (Sec. VIII)\", \"tolerant to small environment changes such as opened or closed doors (Sec. VIII)\"]",[24,25,26],"[\"tested only indoors","whether local normal distributions model less structured or outdoor scenes well is left open (Sec. IX)\", \"mapping without odometry works only while 2D structure is present (Sec. VIII)\", \"full keyframe optimization is not real time as keyframes grow, so only a three-edge subgraph is optimized (Sec. VII-B)\", \"no ground-truth accuracy evaluation","comparison of convergence radius with Lu and Milios left for future work (Sec. VIII, IX)\", \"NDT accuracy is sensitive to voxel resolution (koide2021vgicp Sec. I, secondary, for 3D NDT)\"]",[28],"[\"SICK 2D laser scanner, 180 deg field of view, 1 deg angular resolution (Sec. VIII)\"]",[30],"[\"indoor mobile robot (model not named), driven from the lab along a corridor and back (Sec. VIII)\"]","Newton's method on the negative NDT score (sum of Gaussian evaluations of transformed points), with analytic gradient and Hessian; Hessian replaced by H + lambda I when not positive definite (Sec. IV, V)","no explicit correspondences; points scored against per-cell normal distributions (abstract)","not_applicable (pairwise scan matching)","not_reported","none demonstrated; the authors note that closing a cycle would require optimizing over all keyframes (Sec. VII-B)","local graph optimization: pairwise matching results give quadratic score models (Taylor expansion with the converged Hessian) summed over edges; optimized only over the subgraph within three edges of the new keyframe to stay real time (Sec. VII-B)","per-scan 2D grid of 100 cm cells, each with a normal distribution (mean and covariance, at least three points), using four overlapping grids shifted by half a cell; the map is a collection of keyframes with global poses (Sec. III, VII)","none required; the estimate is initialized by zero, odometry, or linear extrapolation of the previous step; the reported experiments used no odometry (Sec. IV, VI, VIII)","2D pose (tx, ty, phi) per scan; map of 33 keyframes with poses and the estimated trajectory (Fig. 2, Sec. VIII)","building an NDT takes around 10 ms and one Newton iteration around 2 ms on a 1.4 GHz machine, typically 1 to 5 iterations; offline processing of the test run took 58 s, 97 scans per second; Java implementation (Sec. VI, VIII)",null,"not_verified",[],{"id":5,"kind":45,"shortName":7,"title":8,"authors":46,"year":9,"venue":49,"venueType":50,"publisher":51,"volumeIssuePages":52,"doi":53,"arxivId":41,"url":54,"firstPublicDate":55,"publicationStatus":16,"metadataStatus":56,"fulltextStatus":15,"era":10,"classicReason":57,"codeUrl":41,"cluster":11,"topics":58,"mdpi":59,"verification":60,"label":6,"fulltextRoute":61,"versionRead":62,"addedByCensus":59},"method",[47,48],"P. Biber","W. Strasser","Proceedings 2003 IEEE\u002FRSJ International Conference on Intelligent Robots and Systems (IROS 2003)","conference","IEEE","vol. 3, pp. 2743-2748","10.1109\u002Firos.2003.1249285","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1109\u002FIROS.2003.1249285","2003","metadata_verified","principle reused: correspondence-free registration against per-cell normal distributions, generalized to 3D by magnusson2007ndt3d and contrasted by koide2021vgicp.",[11],false,"confirmed","NTU institutional (curl)","version of record, Proc. IEEE\u002FRSJ IROS 2003, vol. 3, pp. 2743-2748, PDF with a text layer (OCR quality moderate; equations and Fig. 1 region checked on the rendered page)",[64,70],{"category":65,"model":66,"canonical":66,"role":67,"dataset":41,"specs":68,"locator":69},"lidar","SICK laser scanner","method input","covering 180 degree with an angular resolution of one degree (model number not stated)","Sec. VIII",{"category":71,"model":72,"canonical":72,"role":73,"dataset":41,"specs":74,"locator":75},"compute","1.4 GHz machine","compute for runtime","Java implementation","Sec. VI, VIII",[],{"totalRows":78,"groupCount":79,"groups":80,"others":145},4,2,[81,117],{"slug":82,"group":83,"sourceId":5,"sourceLabel":6,"table":84,"selfRows":79,"metrics":85,"seqs":92,"entrants":97,"cells":101,"outcomes":108,"locators":109,"hardware":111,"wordings":113,"notes":114},"biber2003ndt-text-sec-vi","biber2003ndt:Text Sec.VI","Text Sec.VI",[86,90],{"label":87,"unit":88,"statistic":34,"alignment":89},"time to build the NDT of a scan (around)","ms","none",{"label":91,"unit":88,"statistic":34,"alignment":89},"time per Newton iteration (around)",[93],{"dataset":94,"sequence":95,"environment":96},"authors' indoor SICK scans","not_applicable","indoor laboratory and corridor",[98],{"name":99,"methodId":5,"linkable":100,"proposed":100,"self":100},"NDT (proposed)",true,[102,106],[103,103,103,104,105,103,103,105,103],0,10,-1,[103,107,103,79,105,103,103,105,107],1,[],[110],"Sec. VI",[112],"1.4 GHz machine; Java implementation",[],[115,116],"Position tracking against a keyframe; per-scan cost of building the NDT","Position tracking against a keyframe; typically 1 to 5 Newton iterations for small movements",{"slug":118,"group":119,"sourceId":5,"sourceLabel":6,"table":120,"selfRows":79,"metrics":121,"seqs":128,"entrants":131,"cells":134,"outcomes":139,"locators":140,"hardware":141,"wordings":142,"notes":143},"biber2003ndt-text-sec-viii","biber2003ndt:Text Sec.VIII","Text Sec.VIII",[122,125],{"label":123,"unit":124,"statistic":34,"alignment":89},"time to process all frames offline","s",{"label":126,"unit":127,"statistic":34,"alignment":89},"scans processed per second offline","scans\u002Fs",[129],{"dataset":94,"sequence":130,"environment":96},"lab to corridor and back, about 83 m, 20 min",[132],{"name":133,"methodId":5,"linkable":100,"proposed":100,"self":100},"NDT scan matcher with keyframe SLAM (proposed)",[135,137],[103,103,103,136,105,103,103,105,103],58,[103,107,103,138,105,103,103,105,103],97,[],[69],[112],[],[144],"Offline processing of the lab-corridor run (every fifth of 28 430 scans used, about 23 scans\u002Fs at a simulated 35 cm\u002Fs; tracking every scan, SLAM step every tenth scan, no odometry)",[],1790510661857]