[{"data":1,"prerenderedAt":480},["ShallowReactive",2],{"method-refusion2019":3},{"method":4,"reference":62,"equipment":86,"figures":110,"results":111},{"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":23,"limitations":29,"sensors":35,"platform":37,"estimator":40,"association":41,"timeModel":42,"deskew":43,"loopClosure":44,"globalOptimization":44,"mapRepresentation":45,"prior":44,"outputGeometry":46,"compute":47,"codeUrl":48,"codeLicense":49,"relatedVersions":50},"refusion2019","Palazzolo et al., 2019","ReFusion","ReFusion: 3D Reconstruction in Dynamic Environments for RGB-D Cameras Exploiting Residuals",2019,"recent","C08","odometry_with_local_mapping","ReFusion 是以 TSDF 為模型的 RGB-D 稠密 SLAM，目標是在有多個移動物體的室內場景中只重建靜態部分。位姿估計不渲染合成視圖，而是把目前影格的點直接帶入 TSDF，以內插得到的符號距離作為殘差，並加入體素色彩的光度誤差。第一次配準後，殘差超過門檻的像素視為動態區域，再以考慮深度的區域成長（flood fill）擴展成遮罩，排除後重新配準並整合。另把相機視錐內確定為空的體素標記為自由空間，之後落在自由空間的量測即視為動態物體而拒絕。方法純幾何、不依賴語意偵測，體素以雜湊配置並在 GPU 平行處理；作者同時發布以動作捕捉提供軌跡、以 Leica BLK360 地面雷射掃描提供靜態場景真值的 Bonn RGB-D 動態資料集。","Dynamic-scene RGB-D TSDF SLAM: direct point-to-TSDF plus voxel-colour photometric tracking, dynamic pixels found from large registration residuals and grown by depth-aware flood fill, and free-space carving that rejects measurements in voxels previously observed empty; class agnostic, GPU voxel hashing, released with the Bonn RGB-D Dynamic Dataset (motion-capture trajectories, TLS static-scene reference).","full_text_reviewed","peer_reviewed_published","main_body","論文未在施工現場測試，實驗為 TUM RGB-D 辦公室動態序列，以及作者在室內測試空間錄製、有人走動、搬箱與玩氣球的動態序列。以配準殘差與自由空間排除動態物體、只保留靜態結構的作法，符合施工中有人員與機具移動時的室內掃描需求。其 Bonn 資料集以 Leica BLK360 地面雷射掃描點雲作為靜態場景真值，並用傾斜與轉動標靶把掃描對齊到動作捕捉座標系（Sec. IV-C），此評估流程可作為工地動態場景掃描驗證的參考（推論）。",[20,21,22],"public_benchmark","independent_reference","controlled_experiment",[24,25,26,27,28],"Tracking on TUM dynamic sequences comparable to StaticFusion, and StaticFusion lost track on walking halfsphere (0.681 m vs 0.104 m) (Table II)","On 24 Bonn sequences the worst ATE (0.571 m) is on par with DynaSLAM N+G (0.575 m) and far below StaticFusion (3.586 m) and DynaSLAM G (1.217 m) (Table III)","Best geometric method on crowded scenes (crowd, crowd2, crowd3) (Table III)","Static model closer to the TLS reference than StaticFusion on crowd3 and removing nonobstructing box (Fig. 11)","Class agnostic; no object tracking, so the number and speed of moving objects are not limited (Sec. I)",[30,31,32,33,34],"Feature-based DynaSLAM (geometric or with segmentation) outperforms it on the TUM dynamic sequences (Table II)","A dynamic object observed at a location that is never revisited stays in the model (walking xyz, Fig. 7)","Invalid (zero) depth prevents free-space reasoning; the virtual-depth workaround from 10 frames adds about 0.3 s delay and assumes nothing is closer than the minimum sensor range (Sec. III-E)","Marking all free voxels in the frustum is memory-inefficient (Sec. III-D)","No loop closure or global optimization (inference from Sec. III)",[36],"RGB-D camera (ASUS Xtion Pro LIVE in the Bonn RGB-D Dynamic Dataset; TUM RGB-D dynamic sequences)",[38,39],"handheld (TUM RGB-D dynamic sequences, camera carried by a cameraman)","not_reported (carrier of the ASUS Xtion in the Bonn dataset is not described)","Frame-to-model direct alignment: points of the current frame are transformed into the TSDF and the interpolated SDF value is the geometric residual, plus a photometric residual against voxel colours (weight 0.025); Levenberg-Marquardt on three coarse-to-fine levels, GPU-parallel; a second registration is run after masking dynamic pixels","Correspondence-free point-to-implicit residuals; dynamic pixels are those whose residual exceeds t = gamma times tau squared (gamma 0.5, tau 0.1 m), grown by depth-aware flood fill (threshold 0.007) and dilation","discrete poses","not_applicable (RGB-D input)","none","TSDF with weight and colour per voxel in dynamically allocated voxel-hashed blocks (1 cm voxels, 0.1 m truncation); voxels seen empty in the camera frustum are marked as free space (SDF set to the truncation distance)","mesh of the static part of the scene, trajectory","GPU-parallel voxel and pixel processing; GPU model and runtime not reported; registration takes up to twice as long as without dynamics handling because of the second pass (Sec. III-C)","https:\u002F\u002Fgithub.com\u002FPRBonn\u002Frefusion","CC BY-NC-SA 3.0 Unported (LICENSE.txt checked; non-commercial)",[51,55,58],{"relation":52,"title":53,"doi_or_url":54},"preprint","ReFusion (arXiv v1 to v3; v3 notes acceptance at IROS 2019)","https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.02082",{"relation":56,"title":57,"doi_or_url":48},"code_release","PRBonn\u002Frefusion",{"relation":59,"title":60,"doi_or_url":61},"dataset","Bonn RGB-D Dynamic Dataset","http:\u002F\u002Fwww.ipb.uni-bonn.de\u002Fdata\u002Frgbd-dynamic-dataset",{"id":5,"kind":63,"shortName":7,"title":8,"authors":64,"year":9,"venue":70,"venueType":71,"publisher":72,"volumeIssuePages":73,"doi":74,"arxivId":75,"url":76,"firstPublicDate":77,"publicationStatus":16,"metadataStatus":78,"fulltextStatus":15,"era":10,"classicReason":79,"codeUrl":48,"cluster":11,"topics":80,"mdpi":81,"verification":82,"label":6,"fulltextRoute":83,"versionRead":84,"addedByCensus":85},"method",[65,66,67,68,69],"Emanuele Palazzolo","Jens Behley","Philipp Lottes","Philippe Giguère","Cyrill Stachniss","2019 IEEE\u002FRSJ International Conference on Intelligent Robots and Systems (IROS)","conference","IEEE","pp. 7855-7862","10.1109\u002Firos40897.2019.8967590","1905.02082","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1109\u002FIROS40897.2019.8967590","2019-05-06","metadata_verified","not_applicable",[11],false,"corrected","arXiv","arXiv 1905.02082v3 (2019-08-29), accepted IROS 2019 version; IEEE version of record not compared",true,[87,95,101,106],{"category":88,"model":89,"canonical":90,"role":91,"dataset":92,"specs":93,"locator":94},"rgbd","ASUS Xtion Pro LIVE","Asus Xtion Pro Live","method input","Bonn RGB-D Dynamic Dataset (this paper)","recorded depth always within the valid sensor range for the Bonn sequences","Sec. IV-B; Sec. III-E",{"category":96,"model":97,"canonical":97,"role":98,"dataset":92,"specs":99,"locator":100},"other","Optitrack Prime 13 motion capture system","reference or ground truth","ground-truth sensor trajectories","Sec. IV-B",{"category":102,"model":103,"canonical":103,"role":98,"dataset":92,"specs":104,"locator":105},"tls_scanner","Leica BLK360","high-resolution point cloud of the static part of the test environment","Sec. IV-B, IV-C; Fig. 12b",{"category":96,"model":107,"canonical":107,"role":98,"dataset":92,"specs":108,"locator":109},"tilt and turn targets","located by both the laser scanner and the motion capture system to align the TLS cloud to the motion-capture frame","Sec. IV-C; Fig. 12c",[],{"totalRows":112,"groupCount":113,"groups":114,"others":479},30,2,[115,401],{"slug":116,"group":117,"sourceId":5,"sourceLabel":6,"table":118,"selfRows":119,"metrics":120,"seqs":126,"entrants":176,"cells":187,"outcomes":395,"locators":396,"hardware":397,"wordings":398,"notes":399},"refusion2019-table-iii","refusion2019:Table III","Table III",24,[121],{"label":122,"unit":123,"statistic":124,"alignment":125},"Absolute Trajectory Error (RMS) [m]","m","RMSE","not_reported",[127,130,132,134,136,138,140,142,144,146,148,150,152,154,156,158,160,162,164,166,168,170,172,174],{"dataset":60,"sequence":128,"environment":129},"balloon","indoor test room with people manipulating boxes, balloons or crowding the camera",{"dataset":60,"sequence":131,"environment":129},"balloon2",{"dataset":60,"sequence":133,"environment":129},"balloon tracking",{"dataset":60,"sequence":135,"environment":129},"balloon tracking2",{"dataset":60,"sequence":137,"environment":129},"crowd",{"dataset":60,"sequence":139,"environment":129},"crowd2",{"dataset":60,"sequence":141,"environment":129},"crowd3",{"dataset":60,"sequence":143,"environment":129},"kidnapping box",{"dataset":60,"sequence":145,"environment":129},"kidnapping box2",{"dataset":60,"sequence":147,"environment":129},"moving no box",{"dataset":60,"sequence":149,"environment":129},"moving no box2",{"dataset":60,"sequence":151,"environment":129},"moving o box",{"dataset":60,"sequence":153,"environment":129},"moving o box2",{"dataset":60,"sequence":155,"environment":129},"person tracking",{"dataset":60,"sequence":157,"environment":129},"person tracking2",{"dataset":60,"sequence":159,"environment":129},"placing no box",{"dataset":60,"sequence":161,"environment":129},"placing no box2",{"dataset":60,"sequence":163,"environment":129},"placing no box3",{"dataset":60,"sequence":165,"environment":129},"placing o box",{"dataset":60,"sequence":167,"environment":129},"removing no box",{"dataset":60,"sequence":169,"environment":129},"removing no box2",{"dataset":60,"sequence":171,"environment":129},"removing o box",{"dataset":60,"sequence":173,"environment":129},"synchronous",{"dataset":60,"sequence":175,"environment":129},"synchronous2",[177,179,182,185],{"name":178,"methodId":5,"linkable":85,"proposed":85,"self":85},"Ours (ReFusion)",{"name":180,"methodId":181,"linkable":85,"proposed":81,"self":81},"SF (StaticFusion)","staticfusion2018",{"name":183,"methodId":184,"linkable":81,"proposed":81,"self":81},"DS (G) (DynaSLAM geometric)",null,{"name":186,"methodId":184,"linkable":81,"proposed":81,"self":81},"DS (N+G) (DynaSLAM neural network + geometric)",[188,192,195,197,200,202,204,206,208,210,212,214,216,218,220,222,224,227,229,231,233,236,238,240,242,245,247,249,251,254,256,258,259,262,264,266,267,270,272,274,276,279,281,283,285,288,290,292,294,297,299,301,302,305,307,309,311,314,316,318,320,323,325,327,329,331,333,335,337,340,342,344,346,349,351,353,355,358,360,361,362,365,367,369,370,373,375,377,379,382,384,386,388,390,391,393],[189,189,189,190,191,189,191,191,189],0,0.175,-1,[193,189,189,194,191,189,191,191,189],1,0.233,[113,189,189,196,191,189,191,191,189],0.05,[198,189,189,199,191,189,191,191,189],3,0.03,[189,189,193,201,191,189,191,191,189],0.254,[193,189,193,203,191,189,191,191,189],0.293,[113,189,193,205,191,189,191,191,189],0.142,[198,189,193,207,191,189,191,191,189],0.029,[189,189,113,209,191,189,191,191,189],0.302,[193,189,113,211,191,189,191,191,189],0.221,[113,189,113,213,191,189,191,191,189],0.156,[198,189,113,215,191,189,191,191,189],0.049,[189,189,198,217,191,189,191,191,189],0.322,[193,189,198,219,191,189,191,191,189],0.366,[113,189,198,221,191,189,191,191,189],0.192,[198,189,198,223,191,189,191,191,189],0.035,[189,189,225,226,191,189,191,191,189],4,0.204,[193,189,225,228,191,189,191,191,189],3.586,[113,189,225,230,191,189,191,191,189],1.065,[198,189,225,232,191,189,191,191,189],0.016,[189,189,234,235,191,189,191,191,189],5,0.155,[193,189,234,237,191,189,191,191,189],0.215,[113,189,234,239,191,189,191,191,189],1.217,[198,189,234,241,191,189,191,191,189],0.031,[189,189,243,244,191,189,191,191,189],6,0.137,[193,189,243,246,191,189,191,191,189],0.168,[113,189,243,248,191,189,191,191,189],0.835,[198,189,243,250,191,189,191,191,189],0.038,[189,189,252,253,191,189,191,191,189],7,0.148,[193,189,252,255,191,189,191,191,189],0.336,[113,189,252,257,191,189,191,191,189],0.026,[198,189,252,207,191,189,191,191,189],[189,189,260,261,191,189,191,191,189],8,0.161,[193,189,260,263,191,189,191,191,189],0.263,[113,189,260,265,191,189,191,191,189],0.033,[198,189,260,223,191,189,191,191,189],[189,189,268,269,191,189,191,191,189],9,0.071,[193,189,268,271,191,189,191,191,189],0.141,[113,189,268,273,191,189,191,191,189],0.317,[198,189,268,275,191,189,191,191,189],0.232,[189,189,277,278,191,189,191,191,189],10,0.179,[193,189,277,280,191,189,191,191,189],0.364,[113,189,277,282,191,189,191,191,189],0.052,[198,189,277,284,191,189,191,191,189],0.039,[189,189,286,287,191,189,191,191,189],11,0.343,[193,189,286,289,191,189,191,191,189],0.331,[113,189,286,291,191,189,191,191,189],0.544,[198,189,286,293,191,189,191,191,189],0.044,[189,189,295,296,191,189,191,191,189],12,0.528,[193,189,295,298,191,189,191,191,189],0.309,[113,189,295,300,191,189,191,191,189],0.589,[198,189,295,263,191,189,191,191,189],[189,189,303,304,191,189,191,191,189],13,0.289,[193,189,303,306,191,189,191,191,189],0.484,[113,189,303,308,191,189,191,191,189],0.714,[198,189,303,310,191,189,191,191,189],0.061,[189,189,312,313,191,189,191,191,189],14,0.463,[193,189,312,315,191,189,191,191,189],0.626,[113,189,312,317,191,189,191,191,189],0.817,[198,189,312,319,191,189,191,191,189],0.078,[189,189,321,322,191,189,191,191,189],15,0.106,[193,189,321,324,191,189,191,191,189],0.125,[113,189,321,326,191,189,191,191,189],0.645,[198,189,321,328,191,189,191,191,189],0.575,[189,189,330,271,191,189,191,191,189],16,[193,189,330,332,191,189,191,191,189],0.177,[113,189,330,334,191,189,191,191,189],0.027,[198,189,330,336,191,189,191,191,189],0.021,[189,189,338,339,191,189,191,191,189],17,0.174,[193,189,338,341,191,189,191,191,189],0.256,[113,189,338,343,191,189,191,191,189],0.327,[198,189,338,345,191,189,191,191,189],0.058,[189,189,347,348,191,189,191,191,189],18,0.571,[193,189,347,350,191,189,191,191,189],0.33,[113,189,347,352,191,189,191,191,189],0.267,[198,189,347,354,191,189,191,191,189],0.255,[189,189,356,357,191,189,191,191,189],19,0.041,[193,189,356,359,191,189,191,191,189],0.136,[113,189,356,232,191,189,191,191,189],[198,189,356,232,191,189,191,191,189],[189,189,363,364,191,189,191,191,189],20,0.111,[193,189,363,366,191,189,191,191,189],0.129,[113,189,363,368,191,189,191,191,189],0.022,[198,189,363,336,191,189,191,191,189],[189,189,371,372,191,189,191,191,189],21,0.222,[193,189,371,374,191,189,191,191,189],0.334,[113,189,371,376,191,189,191,191,189],0.362,[198,189,371,378,191,189,191,191,189],0.291,[189,189,380,381,191,189,191,191,189],22,0.441,[193,189,380,383,191,189,191,191,189],0.446,[113,189,380,385,191,189,191,191,189],0.977,[198,189,380,387,191,189,191,191,189],0.015,[189,189,389,368,191,189,191,191,189],23,[193,189,389,334,191,189,191,191,189],[113,189,389,392,191,189,191,191,189],0.887,[198,189,389,394,191,189,191,191,189],0.009,[],[118],[],[],[400],"Bonn RGB-D Dynamic Dataset (24 highly dynamic scenes, ASUS Xtion Pro LIVE, OptiTrack Prime 13 ground truth); ATE RMS; 'o box' = obstructing box, 'no box' = nonobstructing box; all methods run by the authors with default parameters",{"slug":402,"group":403,"sourceId":5,"sourceLabel":6,"table":404,"selfRows":243,"metrics":405,"seqs":407,"entrants":422,"cells":429,"outcomes":472,"locators":474,"hardware":475,"wordings":476,"notes":477},"refusion2019-table-ii","refusion2019:Table II","Table II",[406],{"label":122,"unit":123,"statistic":124,"alignment":125},[408,412,414,416,418,420],{"dataset":409,"sequence":410,"environment":411},"TUM RGB-D","sitting static","indoor office with seated or walking people, handheld RGB-D",{"dataset":409,"sequence":413,"environment":411},"sitting xyz",{"dataset":409,"sequence":415,"environment":411},"sitting halfsphere",{"dataset":409,"sequence":417,"environment":411},"walking static",{"dataset":409,"sequence":419,"environment":411},"walking xyz",{"dataset":409,"sequence":421,"environment":411},"walking halfsphere",[423,424,425,426,427],{"name":178,"methodId":5,"linkable":85,"proposed":85,"self":85},{"name":180,"methodId":181,"linkable":85,"proposed":81,"self":81},{"name":183,"methodId":184,"linkable":81,"proposed":81,"self":81},{"name":186,"methodId":184,"linkable":81,"proposed":81,"self":81},{"name":428,"methodId":184,"linkable":81,"proposed":81,"self":81},"MF (MaskFusion, values from its paper)",[430,431,433,434,436,437,439,440,441,442,443,445,446,448,450,451,452,453,454,455,456,458,460,462,463,465,466,468,470,471],[189,189,189,394,191,189,191,191,189],[193,189,189,432,191,189,191,191,189],0.014,[113,189,189,394,191,189,191,191,189],[198,189,189,435,191,189,191,191,189],0.007,[225,189,189,336,191,189,191,191,189],[189,189,193,438,191,189,191,191,189],0.04,[193,189,193,284,191,189,191,191,189],[113,189,193,394,191,189,191,191,189],[198,189,193,387,191,189,191,191,189],[225,189,193,241,191,189,191,191,189],[189,189,113,444,191,189,191,191,189],0.11,[193,189,113,357,191,189,191,191,189],[113,189,113,447,191,189,191,191,189],0.017,[198,189,113,449,191,189,191,191,189],0.028,[225,189,113,282,191,189,191,191,189],[189,189,198,447,191,189,191,191,189],[193,189,198,387,191,189,191,191,189],[113,189,198,432,191,189,191,191,189],[198,189,198,435,191,189,191,191,189],[225,189,198,223,191,189,191,191,189],[189,189,225,457,191,189,191,191,189],0.099,[193,189,225,459,191,189,191,191,189],0.093,[113,189,225,461,191,189,191,191,189],0.085,[198,189,225,447,191,189,191,191,189],[225,189,225,464,191,189,191,191,189],0.104,[189,189,234,464,191,189,191,191,189],[193,189,234,467,189,189,191,191,189],0.681,[113,189,234,469,191,189,191,191,189],0.084,[198,189,234,257,191,189,191,191,189],[225,189,234,322,191,189,191,191,189],[473],"other: StaticFusion lost track at the start of this sequence because of excessive dynamic elements (Sec. IV-A)",[404],[],[],[478],"TUM RGB-D dynamic scenes; ATE RMS; ReFusion uses virtual depth from 10 frames (about 0.3 s delay) to fill invalid depth; ReFusion, SF and MF are dense, DynaSLAM is feature-based; StaticFusion lost track on walking halfsphere",[],1790510660221]