[{"data":1,"prerenderedAt":473},["ShallowReactive",2],{"method-manhattanslam2021":3},{"method":4,"reference":58,"equipment":80,"figures":88,"results":89},{"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":27,"sensors":33,"platform":35,"estimator":38,"association":39,"timeModel":40,"deskew":41,"loopClosure":42,"globalOptimization":43,"mapRepresentation":44,"prior":45,"outputGeometry":46,"compute":47,"codeUrl":48,"codeLicense":49,"relatedVersions":50},"manhattanslam2021","Yunus et al., 2021","ManhattanSLAM","ManhattanSLAM: Robust Planar Tracking and Mapping Leveraging Mixture of Manhattan Frames",2021,"recent","C08","odometry_with_local_mapping","ManhattanSLAM 是只用 CPU 的室內 RGB-D SLAM。每一影格擷取 ORB 點、LSD 線段與深度圖中的平面；只要找到兩或三個相互垂直的平面就組成一個曼哈頓座標系（Manhattan Frame），並把場景視為多個曼哈頓座標系的混合。若目前觀測到的曼哈頓座標系先前已存入曼哈頓地圖，就直接由兩次觀測求出無漂移的旋轉，平移再由點、線、面特徵最佳化；若場景不符合曼哈頓假設，則以點、線、面及平行與垂直平面約束估計完整六自由度位姿。稠密建圖沿用 Dense Surfel Mapping 的超像素面元，但平面區域改以稀疏地圖中的平面點建立面元，以節省記憶體。","CPU RGB-D SLAM for structured interiors: points, lines and planes are tracked; Manhattan frames detected from mutually perpendicular planes give drift-free rotation when revisited (translation then from features), with full point-line-plane optimization in non-Manhattan scenes; dense surfels reuse sparse-map planes for planar regions and superpixel surfels elsewhere.","full_text_reviewed","peer_reviewed_published","main_body","論文未在施工現場測試，評估限於 ICL-NUIM 合成室內、TUM RGB-D，以及 TAMU RGB-D 的室內走廊與大廳長序列。以相互垂直平面求無漂移旋轉的作法適合牆、樓板與天花板占多數的建築室內；但未完工或雜亂的施工現場可能不符合曼哈頓假設，此時系統會退回一般特徵追蹤（推論）。",[20,21],"public_benchmark","simulation",[23,24,25,26],"Lowest average ATE RMSE on the eight ICL-NUIM sequences (0.014 m) among the compared feature-based and Manhattan-based systems (Table I)","Keeps tracking on cluttered TUM fr1 and fr2 sequences where the Manhattan-only methods fail (Table I)","Loop-end drift on TAMU RGB-D of 0.53 m and 0.39 m versus 3.13 m and 2.22 m for ORB-SLAM2 without bundle adjustment and loop closure; Manhattan-frame tracking lowers drift from 0.77 to 0.53 m and from 0.81 to 0.39 m compared with feature tracking only (Table II)","Dense reconstruction error 0.5 to 0.7 cm on ICL-NUIM living room, best on three of four sequences, on a CPU (Table III)",[28,29,30,31,32],"No loop-closure module (Sec. V)","Noisy real depth degrades plane normals, so Manhattan-frame tracking is used less and the system falls back to feature tracking (Sec. IV-A2, IV-B)","Planar surfel radius is fixed from the voxel size and does not follow the plane boundary (Sec. V)","TAMU drift is measured only as the start-to-end distance of a loop because no ground truth exists (Sec. IV-B)","ORB-SLAM2 and SP-SLAM were run with bundle adjustment and loop closure disabled, so the comparison excludes their global correction (Sec. IV)",[34],"RGB-D camera (synthetic ICL-NUIM, TUM RGB-D and TAMU RGB-D sequences; sensor models not named in the paper)",[36,37],"real RGB-D sequences (TUM RGB-D; TAMU RGB-D long indoor loop sequences; capture platform not described in the paper)","simulation (ICL-NUIM living room and office)","Keyframe-based feature SLAM with constant-velocity prediction and local-map refinement; in Manhattan scenes rotation is taken drift-free from a previously stored Manhattan-frame observation and only translation is optimized from point, line and plane errors; otherwise full 6-DoF Levenberg-Marquardt with Huber cost over point, line and plane reprojection errors plus parallel and perpendicular plane constraints","ORB points matched by projection and Hamming distance; LSD line segments matched with LBD descriptors; planes extracted by AHC from the downsampled point cloud and matched by normal angle and point-plane distance; Manhattan frames detected from two or three mutually perpendicular plane normals (SVD-orthogonalized) and matched through shared map-plane IDs","discrete poses","not_applicable (RGB-D input)","none (adding a loop-closure module is listed as future work)","none reported; drift is limited by Manhattan-frame rotation estimates","sparse map of point, line and plane landmarks with keyframe co-visibility graph and a Manhattan map of MF observations; dense surfel map in which planar regions reuse sparse-map plane points (surfel radius from the 0.2 m downsampling voxel) and non-planar regions use superpixel surfels as in Dense Surfel Mapping","Manhattan or mixture-of-Manhattan-frames structural assumption where detected; not required","trajectory, sparse point-line-plane map, dense surfel point cloud","CPU only; Intel Core i5-8250U at 1.60 GHz x 8 with 19.5 GB RAM; about 15 Hz, 67 ms tracking and 40 ms superpixel extraction plus surfel fusion on a separate thread (Sec. IV)","https:\u002F\u002Fgithub.com\u002Frazayunus\u002FManhattanSLAM","GPL-3.0 (LICENSE.txt checked; closed-source commercial licence on request)",[51,55],{"relation":52,"title":53,"doi_or_url":54},"preprint","ManhattanSLAM (arXiv v1)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.15068",{"relation":56,"title":57,"doi_or_url":48},"code_release","razayunus\u002FManhattanSLAM (released after the arXiv v1, which states the code will be open-sourced in the future)",{"id":5,"kind":59,"shortName":7,"title":8,"authors":60,"year":9,"venue":64,"venueType":65,"publisher":66,"volumeIssuePages":67,"doi":68,"arxivId":69,"url":70,"firstPublicDate":71,"publicationStatus":16,"metadataStatus":72,"fulltextStatus":15,"era":10,"classicReason":73,"codeUrl":48,"cluster":11,"topics":74,"mdpi":75,"verification":76,"label":6,"fulltextRoute":77,"versionRead":78,"addedByCensus":79},"method",[61,62,63],"Raza Yunus","Yanyan Li","Federico Tombari","2021 IEEE International Conference on Robotics and Automation (ICRA)","conference","IEEE","pp. 6687-6693","10.1109\u002Ficra48506.2021.9562030","2103.15068","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1109\u002FICRA48506.2021.9562030","2021-03-28","metadata_verified","not_applicable",[11],false,"corrected","arXiv","arXiv 2103.15068v1 (2021-03-28), the only arXiv version; ICRA 2021 version of record not compared",true,[81],{"category":82,"model":83,"canonical":83,"role":84,"dataset":85,"specs":86,"locator":87},"compute","Intel Core i5-8250U CPU @ 1.60GHz x 8, 19.5 GB RAM","compute for runtime",null,"no GPU used; about 15 Hz","Sec. IV",[],{"totalRows":90,"groupCount":91,"groups":92,"others":472},29,4,[93,339,378,434],{"slug":94,"group":95,"sourceId":5,"sourceLabel":6,"table":96,"selfRows":97,"metrics":98,"seqs":104,"entrants":145,"cells":159,"outcomes":331,"locators":334,"hardware":335,"wordings":336,"notes":337},"manhattanslam2021-table-i","manhattanslam2021:Table I","Table I",18,[99],{"label":100,"unit":101,"statistic":102,"alignment":103},"ATE RMSE (m)","m","RMSE","not_reported",[105,109,111,113,115,117,119,121,123,127,129,131,133,135,137,139,141,143],{"dataset":106,"sequence":107,"environment":108},"ICL-NUIM","lr-kt0","synthetic living room and office",{"dataset":106,"sequence":110,"environment":108},"lr-kt1",{"dataset":106,"sequence":112,"environment":108},"lr-kt2",{"dataset":106,"sequence":114,"environment":108},"lr-kt3",{"dataset":106,"sequence":116,"environment":108},"of-kt0",{"dataset":106,"sequence":118,"environment":108},"of-kt1",{"dataset":106,"sequence":120,"environment":108},"of-kt2",{"dataset":106,"sequence":122,"environment":108},"of-kt3",{"dataset":124,"sequence":125,"environment":126},"TUM RGB-D","fr1\u002Fxyz","real indoor scenes with varying structure and texture, RGB-D camera (carrying mode not stated in the paper)",{"dataset":124,"sequence":128,"environment":126},"fr1\u002Fdesk",{"dataset":124,"sequence":130,"environment":126},"fr2\u002Fxyz",{"dataset":124,"sequence":132,"environment":126},"fr2\u002Fdesk",{"dataset":124,"sequence":134,"environment":126},"fr3\u002Fs-nt-far",{"dataset":124,"sequence":136,"environment":126},"fr3\u002Fs-nt-near",{"dataset":124,"sequence":138,"environment":126},"fr3\u002Fs-t-near",{"dataset":124,"sequence":140,"environment":126},"fr3\u002Fs-t-far",{"dataset":124,"sequence":142,"environment":126},"fr3\u002Fcabinet",{"dataset":124,"sequence":144,"environment":126},"fr3\u002Fl-cabinet",[146,148,150,152,155,157],{"name":147,"methodId":5,"linkable":79,"proposed":79,"self":79},"Ours (ManhattanSLAM)",{"name":149,"methodId":85,"linkable":75,"proposed":75,"self":75},"S-SLAM [11]",{"name":151,"methodId":85,"linkable":75,"proposed":75,"self":75},"RGBD-SLAM [12] (Li et al. 2020)",{"name":153,"methodId":154,"linkable":79,"proposed":75,"self":75},"ORB-SLAM2 [6] (BA and loop closure disabled)","orbslam2_2017",{"name":156,"methodId":85,"linkable":75,"proposed":75,"self":75},"SP-SLAM [5] (BA and loop closure disabled)",{"name":158,"methodId":85,"linkable":75,"proposed":75,"self":75},"L-SLAM [10]",[160,164,166,169,172,174,177,179,181,182,183,184,186,187,189,191,193,195,197,198,200,202,204,206,208,210,211,213,215,217,218,220,221,222,224,225,226,228,229,230,232,233,235,237,239,240,241,242,243,246,247,248,249,250,251,253,254,255,256,257,258,261,262,263,265,266,267,270,271,272,274,275,276,278,280,281,282,283,285,288,289,290,291,293,295,297,298,299,300,301,303,305,306,307,308,309,311,313,314,316,317,318,320,323,324,326,327,329],[161,161,161,162,163,161,163,163,161],0,0.007,-1,[165,161,161,85,161,161,163,163,161],1,[167,161,161,168,163,161,163,163,161],2,0.006,[170,161,161,171,163,161,163,163,161],3,0.014,[91,161,161,173,163,161,163,163,161],0.019,[175,161,161,176,163,161,163,163,161],5,0.015,[161,161,165,178,163,161,163,163,161],0.011,[165,161,165,180,163,161,163,163,161],0.016,[167,161,165,176,163,161,163,163,161],[170,161,165,178,163,161,163,163,161],[91,161,165,176,163,161,163,163,161],[175,161,165,185,163,161,163,163,161],0.027,[161,161,167,176,163,161,163,163,161],[165,161,167,188,163,161,163,163,161],0.045,[167,161,167,190,163,161,163,163,161],0.02,[170,161,167,192,163,161,163,163,161],0.021,[91,161,167,194,163,161,163,163,161],0.017,[175,161,167,196,163,161,163,163,161],0.053,[161,161,170,178,163,161,163,163,161],[165,161,170,199,163,161,163,163,161],0.046,[167,161,170,201,163,161,163,163,161],0.012,[170,161,170,203,163,161,163,163,161],0.018,[91,161,170,205,163,161,163,163,161],0.022,[175,161,170,207,163,161,163,163,161],0.143,[161,161,91,209,163,161,163,163,161],0.025,[165,161,91,85,161,161,163,163,161],[167,161,91,212,163,161,163,163,161],0.041,[170,161,91,214,163,161,163,163,161],0.049,[91,161,91,216,163,161,163,163,161],0.031,[175,161,91,190,163,161,163,163,161],[161,161,175,219,163,161,163,163,161],0.013,[165,161,175,85,165,161,163,163,161],[167,161,175,190,163,161,163,163,161],[170,161,175,223,163,161,163,163,161],0.029,[91,161,175,203,163,161,163,163,161],[175,161,175,176,163,161,163,163,161],[161,161,227,176,163,161,163,163,161],6,[165,161,227,216,163,161,163,163,161],[167,161,227,178,163,161,163,163,161],[170,161,227,231,163,161,163,163,161],0.03,[91,161,227,185,163,161,163,163,161],[175,161,227,234,163,161,163,163,161],0.026,[161,161,236,219,163,161,163,163,161],7,[165,161,236,238,163,161,163,163,161],0.065,[167,161,236,171,163,161,163,163,161],[170,161,236,201,163,161,163,163,161],[91,161,236,201,163,161,163,163,161],[175,161,236,178,163,161,163,163,161],[161,161,244,245,163,161,163,163,161],8,0.01,[165,161,244,85,165,161,163,163,161],[167,161,244,85,165,161,163,163,161],[170,161,244,245,163,161,163,163,161],[91,161,244,245,163,161,163,163,161],[175,161,244,85,161,161,163,163,161],[161,161,252,185,163,161,163,163,161],9,[165,161,252,85,165,161,163,163,161],[167,161,252,85,165,161,163,163,161],[170,161,252,205,163,161,163,163,161],[91,161,252,234,163,161,163,163,161],[175,161,252,85,161,161,163,163,161],[161,161,259,260,163,161,163,163,161],10,0.008,[165,161,259,85,165,161,163,163,161],[167,161,259,85,165,161,163,163,161],[170,161,259,264,163,161,163,163,161],0.009,[91,161,259,264,163,161,163,163,161],[175,161,259,85,161,161,163,163,161],[161,161,268,269,163,161,163,163,161],11,0.037,[165,161,268,85,165,161,163,163,161],[167,161,268,85,165,161,163,163,161],[170,161,268,273,163,161,163,163,161],0.04,[91,161,268,209,163,161,163,163,161],[175,161,268,85,161,161,163,163,161],[161,161,277,273,163,161,163,163,161],12,[165,161,277,279,163,161,163,163,161],0.281,[167,161,277,205,163,161,163,163,161],[170,161,277,85,165,161,163,163,161],[91,161,277,216,163,161,163,163,161],[175,161,277,284,163,161,163,163,161],0.141,[161,161,286,287,163,161,163,163,161],13,0.023,[165,161,286,238,163,161,163,163,161],[167,161,286,209,163,161,163,163,161],[170,161,286,85,165,161,163,163,161],[91,161,286,292,163,161,163,163,161],0.024,[175,161,286,294,163,161,163,163,161],0.066,[161,161,296,201,163,161,163,163,161],14,[165,161,296,171,163,161,163,163,161],[167,161,296,85,161,161,163,163,161],[170,161,296,178,163,161,163,163,161],[91,161,296,245,163,161,163,163,161],[175,161,296,302,163,161,163,163,161],0.156,[161,161,304,205,163,161,163,163,161],15,[165,161,304,171,163,161,163,163,161],[167,161,304,85,161,161,163,163,161],[170,161,304,178,163,161,163,163,161],[91,161,304,180,163,161,163,163,161],[175,161,304,310,163,161,163,163,161],0.212,[161,161,312,287,163,161,163,163,161],16,[165,161,312,85,161,161,163,163,161],[167,161,312,315,163,161,163,163,161],0.035,[170,161,312,85,165,161,163,163,161],[91,161,312,85,165,161,163,163,161],[175,161,312,319,163,161,163,163,161],0.291,[161,161,321,322,163,161,163,163,161],17,0.083,[165,161,321,85,161,161,163,163,161],[167,161,321,325,163,161,163,163,161],0.071,[170,161,321,85,165,161,163,163,161],[91,161,321,328,163,161,163,163,161],0.074,[175,161,321,330,163,161,163,163,161],0.14,[332,333],"result not available","tracking failure",[96],[],[],[338],"Translation ATE RMSE (m); ORB-SLAM2 and SP-SLAM run without bundle adjustment and loop closure for fairness; 'x' tracking failure, '-' result not available; the number of frames using Manhattan-frame tracking is also listed in the table (not extracted)",{"slug":340,"group":341,"sourceId":5,"sourceLabel":6,"table":342,"selfRows":91,"metrics":343,"seqs":347,"entrants":354,"cells":359,"outcomes":372,"locators":373,"hardware":374,"wordings":375,"notes":376},"manhattanslam2021-table-ii","manhattanslam2021:Table II","Table II",[344],{"label":345,"unit":101,"statistic":103,"alignment":346},"accumulated drift (m)","none",[348,352],{"dataset":349,"sequence":350,"environment":351},"TAMU RGB-D","Corridor-A","indoor corridor and entry hall loops",{"dataset":349,"sequence":353,"environment":351},"Entry-Hall",[355,356,358],{"name":147,"methodId":5,"linkable":79,"proposed":79,"self":79},{"name":357,"methodId":5,"linkable":79,"proposed":75,"self":79},"Ours\u002F-MF (ablation: feature tracking only)",{"name":153,"methodId":154,"linkable":79,"proposed":75,"self":75},[360,362,364,366,368,370],[161,161,161,361,163,161,163,163,161],0.53,[165,161,161,363,163,161,163,163,161],0.77,[167,161,161,365,163,161,163,163,161],3.13,[161,161,165,367,163,161,163,163,161],0.39,[165,161,165,369,163,161,163,163,161],0.81,[167,161,165,371,163,161,163,163,161],2.22,[],[342],[],[],[377],"TAMU RGB-D long indoor loops without ground truth; drift = Euclidean distance between start and end of the estimated loop trajectory",{"slug":379,"group":380,"sourceId":5,"sourceLabel":6,"table":381,"selfRows":91,"metrics":382,"seqs":386,"entrants":392,"cells":402,"outcomes":428,"locators":429,"hardware":430,"wordings":431,"notes":432},"manhattanslam2021-table-iii","manhattanslam2021:Table III","Table III",[383],{"label":384,"unit":385,"statistic":103,"alignment":103},"reconstruction error (cm)","cm",[387,389,390,391],{"dataset":106,"sequence":107,"environment":388},"synthetic living room",{"dataset":106,"sequence":110,"environment":388},{"dataset":106,"sequence":112,"environment":388},{"dataset":106,"sequence":114,"environment":388},[393,396,398,401],{"name":394,"methodId":395,"linkable":79,"proposed":75,"self":75},"E-Fus [15] (ElasticFusion, IJRR version)","elasticfusion2015",{"name":397,"methodId":85,"linkable":75,"proposed":75,"self":75},"InfiniTAM [41] (InfiniTAM v3 report)",{"name":399,"methodId":400,"linkable":79,"proposed":75,"self":75},"DSM [14] (Dense Surfel Mapping)","densesurfelmapping2019",{"name":147,"methodId":5,"linkable":79,"proposed":79,"self":79},[403,405,407,408,410,411,413,415,417,419,421,422,423,425,426,427],[161,161,161,404,163,161,163,163,161],0.7,[165,161,161,406,163,161,163,163,161],1.3,[167,161,161,404,163,161,163,163,161],[170,161,161,409,163,161,163,163,161],0.5,[161,161,165,404,163,161,163,163,161],[165,161,165,412,163,161,163,163,161],1.1,[167,161,165,414,163,161,163,163,161],0.9,[170,161,165,416,163,161,163,163,161],0.6,[161,161,167,418,163,161,163,163,161],0.8,[165,161,167,420,163,161,163,163,161],0.1,[167,161,167,412,163,161,163,163,161],[170,161,167,404,163,161,163,163,161],[161,161,170,424,163,161,163,163,161],2.8,[165,161,170,424,163,161,163,163,161],[167,161,170,165,163,161,163,163,161],[170,161,170,404,163,161,163,163,161],[],[381],[],[],[433],"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":435,"group":436,"sourceId":5,"sourceLabel":6,"table":437,"selfRows":170,"metrics":438,"seqs":448,"entrants":452,"cells":458,"outcomes":464,"locators":466,"hardware":467,"wordings":469,"notes":470},"manhattanslam2021-text-sec-iv","manhattanslam2021:Text Sec.IV","Text Sec.IV",[439,443,446],{"label":440,"unit":441,"statistic":442,"alignment":103},"runs at around 15 Hz","Hz","mean",{"label":444,"unit":445,"statistic":442,"alignment":103},"tracking time","ms",{"label":447,"unit":445,"statistic":442,"alignment":103},"superpixel extraction and surfel fusion time",[449],{"dataset":450,"sequence":103,"environment":451},"ICL-NUIM, TUM RGB-D and TAMU RGB-D","indoor",[453,454,456],{"name":147,"methodId":5,"linkable":79,"proposed":79,"self":79},{"name":455,"methodId":5,"linkable":79,"proposed":79,"self":79},"Ours (ManhattanSLAM, tracking)",{"name":457,"methodId":5,"linkable":79,"proposed":79,"self":79},"Ours (ManhattanSLAM, dense mapping thread)",[459,460,462],[161,161,161,304,161,161,161,163,161],[165,165,161,461,163,161,161,163,161],67,[167,167,161,463,163,161,161,163,161],40,[465],"other: approximate value ('around 15 Hz', Sec. IV)",[87],[468],"Intel Core i5-8250U @ 1.60 GHz x 8, 19.5 GB RAM, no GPU",[],[471],"Average timing over the experiments; dense mapping runs on a separate thread",[],1790510664194]