[{"data":1,"prerenderedAt":287},["ShallowReactive",2],{"method-densesurfelmapping2019":3},{"method":4,"reference":62,"equipment":84,"figures":109,"results":110},{"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":28,"sensors":33,"platform":37,"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},"densesurfelmapping2019","Wang et al., 2019","Dense Surfel Mapping (Wang, Gao, Shen)","Real-time Scalable Dense Surfel Mapping",2019,"recent","C08","map_representation_or_reconstruction","作者提出只用 CPU 的面元（surfel）稠密建圖系統，相機位姿、參考關鍵影格與位姿圖都由外部稀疏視覺 SLAM（ORB-SLAM2 或 VINS-Mono）提供。每張影像先以擴充的 SLIC 依強度、深度與位置分割超像素，可處理無效深度，並以 Huber 損失求穩健的平均深度，再由超像素建立面元，因此能融合 RGB-D、立體相機或單眼深度預測等品質較差的深度圖。每個面元附屬於一個關鍵影格，只有在位姿圖上與參考關鍵影格相距少於 G_delta 條邊的面元才參與融合，使每一影格的融合時間與場景規模無關；位姿圖最佳化後，依各關鍵影格的位姿變化移動其面元，使整張地圖非剛性變形並保持全域一致。","CPU surfel fusion driven by a sparse SLAM pose graph: superpixel surfels (SLIC extended to intensity and depth with Huber-robust depth) are attached to keyframes, only surfels of pose-graph-local keyframes are fused so per-frame cost stays constant, and loop closures deform the map by moving each keyframe's surfels with its optimized pose; works with RGB-D, stereo or monocular predicted depth.","full_text_reviewed","peer_reviewed_published","main_body","原論文未在施工現場測試。corpus 中 chen2025quadrupedinspection 在 Unitree Go1 四足機器人上，以 Intel RealSense D455 的 RGB-D 資料執行本方法，對香港科技大學既有建築約 500 平方公尺的室內區域建圖，並以 Leica BLK360 地面雷射掃描為參考，經 ICP 對齊後的單向平均最近鄰距離為 0.1193 m、標準差 0.1356 m（該文 Table 8、Sec. 3.3.2）。該數值評估的是整體定位加建圖流程，並非單獨評估本方法。",[20,21],"public_benchmark","simulation",[23,24,25,26,27],"Mean reconstruction error 0.7 to 1.1 cm on ICL-NUIM kt0 to kt3 with CPU only, similar to FlashFusion (Table I)","Loop closure reduces kt3 error from 1.7 cm to 0.8 cm (Table I; Sec. VI-A)","Real-time urban-scale reconstruction of KITTI 00 on CPU: about 80 ms per frame, surfel fusion under 6 ms regardless of scale (Sec. VI-B, Fig. 6)","Memory grows with environment size rather than runtime because surfels are reused on revisits (Fig. 7)","Map deformation keeps revisited fine obstacles consistent, unlike CHISEL given the same images and poses (Sec. VI-D, Fig. 8)",[29,30,31,32],"Depends on an external sparse SLAM system for poses, loop detection and pose-graph optimization (Sec. III, IV-B)","Assumes keyframes within G_delta edges are locally consistent and moves surfels rigidly with their keyframe (Sec. IV-B, IV-C)","Large-scale and flight experiments are evaluated by runtime, memory and qualitative figures only, without geometric ground truth (Sec. VI-B to VI-D)","Most computation goes to superpixel extraction and surfel initialization rather than fusion (Sec. VI-B)",[34,35,36],"RGB-D (synthetic ICL-NUIM input with ORB-SLAM2 in RGB-D mode)","Stereo camera (KITTI odometry; depth from PSMNet stereo matching, ORB-SLAM2 stereo mode)","Monocular camera with learned depth (KITTI left images with monocular depth prediction; handheld camera with MVDepthNet depth and VINS-Mono tracking)",[38,39,40,41],"vehicle (KITTI odometry 00 and 05)","handheld (monocular camera used to build the map for quadrotor flights)","UAV (quadrotor flew aggressive autonomous paths planned on the reconstructed map)","simulation (ICL-NUIM synthetic room)","No pose estimation in the mapper: an external sparse visual SLAM (ORB-SLAM2 or VINS-Mono) provides each frame's pose, the reference keyframe and the optimized keyframe pose graph","Local surfels are projected into the current frame and matched to the surfel initialized from the superpixel at that pixel when depths agree within a disparity-derived bound and normals agree (dot product > 0.8); only surfels attached to keyframes within G_delta pose-graph edges of the reference keyframe (breadth-first search) are fused","discrete keyframe poses","not_applicable (camera input)","Provided by the localization system (ORB-SLAM2 or VINS-Mono); after loop closure previously built surfels are reused through new pose-graph edges","Pose graph optimized by the external SLAM; surfels are deformed by applying each keyframe's pose change to its attached surfels (rigid per keyframe, non-rigid overall)","superpixel-based surfels (position, normal, intensity, weight, radius, update count, attached keyframe index) stored in a map database organized by keyframe","none","globally consistent surfel map (shown as point clouds and meshes)","CPU only; workstation with Intel i7-7700; average fusion time about 80 ms per frame on KITTI 00 (above 10 Hz), of which surfel fusion takes less than 6 ms regardless of scale; superpixel grid spacing 8 pixels","https:\u002F\u002Fgithub.com\u002FHKUST-Aerial-Robotics\u002FDenseSurfelMapping","not_stated (no LICENSE file; package.xml licence field reads 'TODO')",[55,59],{"relation":56,"title":57,"doi_or_url":58},"preprint","Real-time Scalable Dense Surfel Mapping (arXiv v1, posted after ICRA 2019)","https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.04250",{"relation":60,"title":61,"doi_or_url":52},"code_release","HKUST-Aerial-Robotics\u002FDenseSurfelMapping",{"id":5,"kind":63,"shortName":7,"title":8,"authors":64,"year":9,"venue":68,"venueType":69,"publisher":70,"volumeIssuePages":71,"doi":72,"arxivId":73,"url":74,"firstPublicDate":75,"publicationStatus":16,"metadataStatus":76,"fulltextStatus":15,"era":10,"classicReason":77,"codeUrl":52,"cluster":11,"topics":78,"mdpi":79,"verification":80,"label":6,"fulltextRoute":81,"versionRead":82,"addedByCensus":83},"method",[65,66,67],"Kaixuan Wang","Fei Gao","Shaojie Shen","2019 International Conference on Robotics and Automation (ICRA)","conference","IEEE","pp. 6919-6925","10.1109\u002Ficra.2019.8794101","1909.04250","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1109\u002FICRA.2019.8794101","2019-05-20","metadata_verified","not_applicable",[11],false,"corrected","arXiv","arXiv 1909.04250v1 (2019-09-10), stated as the ICRA 2019 paper; IEEE version of record not compared",true,[85,92,99,105],{"category":86,"model":87,"canonical":87,"role":88,"dataset":89,"specs":90,"locator":91},"compute","workstation with an Intel i7-7700","compute for runtime",null,"all mapping on CPU","Sec. VI",{"category":93,"model":94,"canonical":94,"role":95,"dataset":96,"specs":97,"locator":98},"stereo_camera","KITTI stereo cameras (model not reported in the paper)","dataset sensor","KITTI odometry","depth maps from PSMNet stereo matching; left images also used for monocular depth prediction","Sec. VI-B, VI-C",{"category":100,"model":101,"canonical":101,"role":102,"dataset":89,"specs":103,"locator":104},"camera","handheld monocular camera (model not reported)","method input","depth from MVDepthNet, poses from VINS-Mono; map used for quadrotor flights","Sec. VI-D",{"category":106,"model":107,"canonical":107,"role":102,"dataset":89,"specs":108,"locator":104},"platform","quadrotor (model not reported)","flew aggressive autonomous paths planned on the map reconstructed from the handheld monocular camera sequence (Sec. VI-D)",[],{"totalRows":111,"groupCount":112,"groups":113,"others":286},14,3,[114,199,250],{"slug":115,"group":116,"sourceId":5,"sourceLabel":6,"table":117,"selfRows":118,"metrics":119,"seqs":125,"entrants":136,"cells":152,"outcomes":193,"locators":194,"hardware":195,"wordings":196,"notes":197},"densesurfelmapping2019-table-i","densesurfelmapping2019:Table I","Table I",8,[120],{"label":121,"unit":122,"statistic":123,"alignment":124},"reconstruction accuracy (cm)","cm","mean","not_reported",[126,130,132,134],{"dataset":127,"sequence":128,"environment":129},"ICL-NUIM","kt0","synthetic indoor living room",{"dataset":127,"sequence":131,"environment":129},"kt1",{"dataset":127,"sequence":133,"environment":129},"kt2",{"dataset":127,"sequence":135,"environment":129},"kt3",[137,140,143,145,148,150],{"name":138,"methodId":139,"linkable":83,"proposed":79,"self":79},"BundleFusion","bundlefusion2017",{"name":141,"methodId":142,"linkable":83,"proposed":79,"self":79},"ElasticFusion","elasticfusion2015",{"name":144,"methodId":89,"linkable":79,"proposed":79,"self":79},"InfiniTAM [13] (Kaehler et al. ECCV 2016)",{"name":146,"methodId":147,"linkable":83,"proposed":79,"self":79},"FlashFusion","flashfusion2018",{"name":149,"methodId":5,"linkable":83,"proposed":83,"self":83},"Ours",{"name":151,"methodId":5,"linkable":83,"proposed":79,"self":83},"Ours w\u002Fo loop (ablation: ORB-SLAM2 loop closure disabled)",[153,157,160,163,165,167,169,171,172,174,175,177,178,179,180,182,183,184,185,186,188,189,190,191],[154,154,154,155,156,154,156,156,154],0,0.5,-1,[158,154,154,159,156,154,156,156,154],1,0.7,[161,154,154,162,156,154,156,156,154],2,1.3,[112,154,154,164,156,154,156,156,154],0.8,[166,154,154,159,156,154,156,156,154],4,[168,154,154,159,156,154,156,156,154],5,[154,154,158,170,156,154,156,156,154],0.6,[158,154,158,159,156,154,156,156,154],[161,154,158,173,156,154,156,156,154],1.1,[112,154,158,164,156,154,156,156,154],[166,154,158,176,156,154,156,156,154],0.9,[168,154,158,176,156,154,156,156,154],[154,154,161,159,156,154,156,156,154],[158,154,161,164,156,154,156,156,154],[161,154,161,181,156,154,156,156,154],0.1,[112,154,161,158,156,154,156,156,154],[166,154,161,173,156,154,156,156,154],[168,154,161,173,156,154,156,156,154],[154,154,112,164,156,154,156,156,154],[158,154,112,187,156,154,156,156,154],2.8,[161,154,112,187,156,154,156,156,154],[112,154,112,162,156,154,156,156,154],[166,154,112,164,156,154,156,156,154],[168,154,112,192,156,154,156,156,154],1.7,[],[117],[],[],[198],"ICL-NUIM living room with simulated noise; reconstruction accuracy = mean difference between reconstructed model and ground-truth model; ORB-SLAM2 RGB-D mode for tracking, G_delta = 20; only FlashFusion and Ours run without GPU; 'Ours w\u002Fo loop' disables ORB-SLAM2 loop closure; comparator values coincide with FlashFusion Table III",{"slug":200,"group":201,"sourceId":202,"sourceLabel":203,"table":204,"selfRows":166,"metrics":205,"seqs":208,"entrants":218,"cells":227,"outcomes":244,"locators":245,"hardware":246,"wordings":247,"notes":248},"manhattanslam2021-table-iii","manhattanslam2021:Table III","manhattanslam2021","Yunus et al., 2021","Table III",[206],{"label":207,"unit":122,"statistic":124,"alignment":124},"reconstruction error (cm)",[209,212,214,216],{"dataset":127,"sequence":210,"environment":211},"lr-kt0","synthetic living room",{"dataset":127,"sequence":213,"environment":211},"lr-kt1",{"dataset":127,"sequence":215,"environment":211},"lr-kt2",{"dataset":127,"sequence":217,"environment":211},"lr-kt3",[219,221,223,225],{"name":220,"methodId":142,"linkable":83,"proposed":79,"self":79},"E-Fus [15] (ElasticFusion, IJRR version)",{"name":222,"methodId":89,"linkable":79,"proposed":79,"self":79},"InfiniTAM [41] (InfiniTAM v3 report)",{"name":224,"methodId":5,"linkable":83,"proposed":79,"self":83},"DSM [14] (Dense Surfel Mapping)",{"name":226,"methodId":202,"linkable":83,"proposed":83,"self":79},"Ours (ManhattanSLAM)",[228,229,230,231,232,233,234,235,236,237,238,239,240,241,242,243],[154,154,154,159,156,154,156,156,154],[158,154,154,162,156,154,156,156,154],[161,154,154,159,156,154,156,156,154],[112,154,154,155,156,154,156,156,154],[154,154,158,159,156,154,156,156,154],[158,154,158,173,156,154,156,156,154],[161,154,158,176,156,154,156,156,154],[112,154,158,170,156,154,156,156,154],[154,154,161,164,156,154,156,156,154],[158,154,161,181,156,154,156,156,154],[161,154,161,173,156,154,156,156,154],[112,154,161,159,156,154,156,156,154],[154,154,112,187,156,154,156,156,154],[158,154,112,187,156,154,156,156,154],[161,154,112,158,156,154,156,156,154],[112,154,112,159,156,154,156,156,154],[],[204],[],[],[249],"ICL-NUIM living room; reconstruction error of the point cloud generated from the surfels (cm); ElasticFusion and InfiniTAM need a GPU, DSM and ManhattanSLAM run on CPU; comparator provenance not stated",{"slug":251,"group":252,"sourceId":5,"sourceLabel":6,"table":253,"selfRows":161,"metrics":254,"seqs":261,"entrants":265,"cells":269,"outcomes":274,"locators":277,"hardware":280,"wordings":282,"notes":283},"densesurfelmapping2019-text-sec-vi-b","densesurfelmapping2019:Text Sec.VI-B","Text Sec.VI-B",[255,258],{"label":256,"unit":257,"statistic":123,"alignment":124},"average fusion time per frame","ms",{"label":259,"unit":257,"statistic":260,"alignment":124},"surfel fusion only consumes less than 6 ms","max",[262],{"dataset":96,"sequence":263,"environment":264},"00","urban driving, vehicle stereo",[266,267],{"name":149,"methodId":5,"linkable":83,"proposed":83,"self":83},{"name":268,"methodId":5,"linkable":83,"proposed":83,"self":83},"Ours (surfel fusion step)",[270,272],[154,154,154,271,154,154,154,156,154],80,[158,158,154,273,158,158,154,156,158],6,[275,276],"other: approximate value ('around 80 ms' per frame, Sec. VI-B)","other: upper bound ('less than 6 ms', Sec. VI-B)",[278,279],"Sec. VI-B; Fig. 6","Sec. VI-B",[281],"Intel i7-7700 workstation (CPU only)",[],[284,285],"KITTI odometry 00 reconstructed from PSMNet stereo depth with ORB-SLAM2 stereo tracking; per-frame fusion time including superpixel extraction and surfel initialization","KITTI odometry 00; time of the surfel fusion step alone, stated as an upper bound independent of environment scale",[],1790510657936]