[{"data":1,"prerenderedAt":429},["ShallowReactive",2],{"method-psmslam2017":3},{"method":4,"reference":50,"equipment":71,"figures":78,"results":79},{"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":32,"platform":34,"estimator":37,"association":38,"timeModel":39,"deskew":40,"loopClosure":41,"globalOptimization":42,"mapRepresentation":43,"prior":44,"outputGeometry":45,"compute":46,"codeUrl":47,"codeLicense":48,"relatedVersions":49},"psmslam2017","Yan et al., 2017","PSM SLAM (Probabilistic Surfel Map)","Dense Visual SLAM with Probabilistic Surfel Map",2017,"recent","C08","full_slam_with_global_correction","PSM SLAM 以機率面元地圖（Probabilistic Surfel Map）結合逐影格與對模型兩類 RGB-D 視覺 SLAM。地圖中每個點帶有三維位置與 3×3 共變異、強度與其變異量以及法向，新觀測以兩個高斯分布相乘的方式融合，正確關聯的點其不確定度會快速下降，不可靠的點則被移除，因此地圖只保留稀疏而可靠的點。前端沿用 σ-DVO 的光度與幾何混合權重，把每一影格對齊到由全域地圖可見點與新關鍵影格觀測組成的 Keyframe PSM；後端以 g2o 交替最佳化關鍵影格之間的位姿約束，以及以不確定度加權的位姿對地圖點約束（光度式光束法平差）。需要稠密網格時，再以地圖點為控制點調整各關鍵影格的深度圖後融合。","Keyframe dense RGB-D SLAM with a Probabilistic Surfel Map: sparse points carrying position covariance and intensity variance are fused by Gaussian products; frames are aligned to a keyframe PSM with sigma-DVO weighting, and the back end alternates pose-pose graph optimization with uncertainty-weighted pose-point (photometric BA) constraints; dense meshes are produced on demand by deforming keyframe depth maps toward the PSM.","full_text_reviewed","peer_reviewed_published","supplementary","論文未在施工現場測試，評估限於 TUM RGB-D 辦公室序列與 ICL-NUIM 合成客廳，目標應用為擴增實境。以不確定度加權的稀疏地圖在 CPU 上降低計算量，並可於需要時輸出稠密點雲，對室內擴增實境檢核有參考價值；但作者指出大量移動物體會使方法失效，且本次查核未找到公開程式碼（推論）。",[20,21],"public_benchmark","simulation",[23,24,25,26],"Visual odometry on 11 TUM sequences: average ATE 0.088 m and RPE 0.034 m\u002Fs versus 0.161 m and 0.042 m\u002Fs for sigma-DVO (Table 1)","Full SLAM ATE lowest among the compared systems on fr1\u002Fdesk (0.016 m), fr1\u002F360 (0.055 m) and fr1\u002Froom (0.051 m) (Table 2)","ICL-NUIM living room: lowest average ATE (0.024 m versus 0.034 m for ElasticFusion) and average surface error equal to ElasticFusion (0.012 m) (Tables 4 and 5)","Sparse sampling (20%) cuts computation time by 30 to 40% compared with DVO and sigma-DVO on the same hardware (Sec. 6.2)",[28,29,30,31],"Can be confused and follow a wrong motion when a large part of the scene moves independently of the camera, for example a hallway with a moving crowd (Sec. 7)","The sparse PSM leaves blank pixels; dense output requires a separate keyframe deformation and fusion step with higher error than the PSM alone (Sec. 6.3; Fig. 6)","Less accurate than sigma-DVO SLAM on fr3\u002Foffice (0.031 m versus 0.015 m) (Tables 2 and 3)","The pose-point map optimization is run at the end of SLAM rather than continuously (Sec. 3 system overview; Sec. 5.2)",[33],"RGB-D camera (TUM RGB-D real sequences; ICL-NUIM synthetic sequences with noise; sensor models not named)",[35,36],"real RGB-D sequences (TUM RGB-D; capture platform not described in the paper)","simulation (ICL-NUIM living room with noise)","Keyframe-based dense visual odometry with the sigma-DVO hybrid weighting (Student-t for photometric, sensor-noise model for geometric residuals), aligning each frame to a Keyframe PSM; back end in g2o that alternates pose-pose graph optimization with uncertainty-weighted pose-point constraints (photometric bundle adjustment on PSM points), usually converging within 10 iterations","PSM points projected into the keyframe define active points; the Keyframe PSM merges them with back-projected new observations (pruned by position uncertainty and sampled at 20%); frame residuals are photometric and depth differences at projected points","discrete poses (keyframes and frames attached to them)","not_applicable (RGB-D input)","Nearest-neighbour search of keyframe poses in a kd-tree (as in sigma-DVO) adds pose-pose constraints","g2o graph optimization alternating pose-pose and pose-point (PSM) constraints at the end of SLAM, with outlier-edge removal","Probabilistic Surfel Map: sparse points with 3D position and 3x3 covariance, intensity and intensity variance, and normal; updated by multiplying Gaussian distributions so that uncertainties of consistently associated points shrink; unreliable points pruned","none","sparse globally consistent PSM; dense point cloud or mesh on request by deforming each keyframe depth map toward nearby PSM points (Gaussian-weighted KNN) and fusing","CPU, multi-threaded, built on the open-source C++ DVO SLAM; workstation with Intel Xeon E5 @ 2.4 GHz; 20% point sample rate gives 30 to 40% less computation time than DVO and sigma-DVO on the same hardware (Sec. 6, 6.2)",null,"not_applicable (no public code found)",[],{"id":5,"kind":51,"shortName":7,"title":8,"authors":52,"year":9,"venue":56,"venueType":57,"publisher":58,"volumeIssuePages":59,"doi":60,"arxivId":47,"url":61,"firstPublicDate":62,"publicationStatus":16,"metadataStatus":63,"fulltextStatus":15,"era":10,"classicReason":64,"codeUrl":47,"cluster":11,"topics":65,"mdpi":66,"verification":67,"label":6,"fulltextRoute":68,"versionRead":69,"addedByCensus":70},"method",[53,54,55],"Zhixin Yan","Mao Ye","Liu Ren","IEEE Transactions on Visualization and Computer Graphics (ISMAR 2017 special issue)","journal","IEEE","23(11):2389-2398","10.1109\u002Ftvcg.2017.2734458","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1109\u002FTVCG.2017.2734458","2017-08-10","metadata_verified","not_applicable",[11],false,"corrected","NTU institutional (Chrome)","IEEE Xplore HTML full text of the version of record (TVCG 23(11), 2017) with Tables 1 to 5 viewed as publisher images",true,[72],{"category":73,"model":74,"canonical":74,"role":75,"dataset":47,"specs":76,"locator":77},"compute","workstation with Intel Xeon E5 @ 2.4GHz","compute for runtime","multi-threaded CPU implementation built on the open-source DVO SLAM; GPU acceleration is mentioned only as a possible improvement (Sec. 6)","Sec. 6",[],{"totalRows":80,"groupCount":81,"groups":82,"others":423},42,5,[83,227,278,354],{"slug":84,"group":85,"sourceId":5,"sourceLabel":6,"table":86,"selfRows":87,"metrics":88,"seqs":97,"entrants":122,"cells":127,"outcomes":221,"locators":222,"hardware":223,"wordings":224,"notes":225},"psmslam2017-table-1","psmslam2017:Table 1","Table 1",22,[89,94],{"label":90,"unit":91,"statistic":92,"alignment":93},"ATE [m]","m","RMSE","not_reported",{"label":95,"unit":96,"statistic":92,"alignment":93},"RPE [m\u002Fs]","m\u002Fs",[98,102,104,106,108,110,112,114,116,118,120],{"dataset":99,"sequence":100,"environment":101},"TUM RGB-D","fr1\u002F360","indoor office scenes, RGB-D camera (carrying mode not stated in the paper)",{"dataset":99,"sequence":103,"environment":101},"fr1\u002Fdesk",{"dataset":99,"sequence":105,"environment":101},"fr1\u002Fdesk2",{"dataset":99,"sequence":107,"environment":101},"fr1\u002Ffloor",{"dataset":99,"sequence":109,"environment":101},"fr1\u002Froom",{"dataset":99,"sequence":111,"environment":101},"fr1\u002Frpy",{"dataset":99,"sequence":113,"environment":101},"fr1\u002Fxyz",{"dataset":99,"sequence":115,"environment":101},"fr2\u002Fdesk",{"dataset":99,"sequence":117,"environment":101},"fr2\u002Frpy",{"dataset":99,"sequence":119,"environment":101},"fr2\u002Fxyz",{"dataset":99,"sequence":121,"environment":101},"fr3\u002Foffice",[123,125],{"name":124,"methodId":47,"linkable":66,"proposed":66,"self":66},"sigma-DVO (visual odometry)",{"name":126,"methodId":5,"linkable":70,"proposed":70,"self":70},"PSM VO (visual odometry)",[128,132,135,137,139,142,144,147,149,152,153,155,157,160,162,165,167,170,172,174,176,179,181,183,185,187,189,191,193,195,197,199,201,203,205,207,209,211,212,214,216,218,219,220],[129,129,129,130,131,129,131,131,129],0,0.229,-1,[133,129,129,134,131,129,131,131,129],1,0.113,[129,129,133,136,131,129,131,131,129],0.067,[133,129,133,138,131,129,131,131,129],0.033,[129,129,140,141,131,129,131,131,129],2,0.088,[133,129,140,143,131,129,131,131,129],0.071,[129,129,145,146,131,129,131,131,129],3,0.226,[133,129,145,148,131,129,131,131,129],0.24,[129,129,150,151,131,129,131,131,129],4,0.314,[133,129,150,130,131,129,131,131,129],[129,129,81,154,131,129,131,131,129],0.072,[133,129,81,156,131,129,131,131,129],0.049,[129,129,158,159,131,129,131,131,129],6,0.052,[133,129,158,161,131,129,131,131,129],0.055,[129,129,163,164,131,129,131,131,129],7,0.184,[133,129,163,166,131,129,131,131,129],0.089,[129,129,168,169,131,129,131,131,129],8,0.188,[133,129,168,171,131,129,131,131,129],0.014,[129,129,173,169,131,129,131,131,129],9,[133,129,173,175,131,129,131,131,129],0.015,[129,129,177,178,131,129,131,131,129],10,0.164,[133,129,177,180,131,129,131,131,129],0.057,[129,133,129,182,131,129,131,131,129],0.11,[133,133,129,184,131,129,131,131,129],0.066,[129,133,133,186,131,129,131,131,129],0.039,[133,133,133,188,131,129,131,131,129],0.028,[129,133,140,190,131,129,131,131,129],0.065,[133,133,140,192,131,129,131,131,129],0.5,[129,133,145,194,131,129,131,131,129],0.053,[133,133,145,196,131,129,131,131,129],0.086,[129,133,150,198,131,129,131,131,129],0.063,[133,133,150,200,131,129,131,131,129],0.05,[129,133,81,202,131,129,131,131,129],0.046,[133,133,81,204,131,129,131,131,129],0.038,[129,133,158,206,131,129,131,131,129],0.036,[133,133,158,208,131,129,131,131,129],0.026,[129,133,163,210,131,129,131,131,129],0.016,[133,133,163,175,131,129,131,131,129],[129,133,168,213,131,129,131,131,129],0.012,[133,133,168,215,131,129,131,131,129],0.005,[129,133,173,217,131,129,131,131,129],0.01,[133,133,173,215,131,129,131,131,129],[129,133,177,171,131,129,131,131,129],[133,133,177,217,131,129,131,131,129],[],[86],[],[],[226],"TUM RGB-D visual odometry only (no back-end); RMSE computed with the TUM benchmark scripts; PSM VO RPE on fr1\u002Fdesk2 printed as 0.50 (likely 0.050)",{"slug":228,"group":229,"sourceId":5,"sourceLabel":6,"table":230,"selfRows":168,"metrics":231,"seqs":233,"entrants":242,"cells":247,"outcomes":272,"locators":273,"hardware":274,"wordings":275,"notes":276},"psmslam2017-table-3","psmslam2017:Table 3","Table 3",[232],{"label":90,"unit":91,"statistic":92,"alignment":93},[234,235,236,237,238,239,240,241],{"dataset":99,"sequence":113,"environment":101},{"dataset":99,"sequence":111,"environment":101},{"dataset":99,"sequence":103,"environment":101},{"dataset":99,"sequence":105,"environment":101},{"dataset":99,"sequence":109,"environment":101},{"dataset":99,"sequence":100,"environment":101},{"dataset":99,"sequence":115,"environment":101},{"dataset":99,"sequence":121,"environment":101},[243,245],{"name":244,"methodId":47,"linkable":66,"proposed":66,"self":66},"sigma-DVO SLAM",{"name":246,"methodId":5,"linkable":70,"proposed":70,"self":70},"PSM SLAM",[248,249,251,252,254,256,257,259,260,262,263,265,266,267,269,270],[129,129,129,210,131,129,131,131,129],[133,129,129,250,131,129,131,131,129],0.011,[129,129,133,188,131,129,131,131,129],[133,129,133,253,131,129,131,131,129],0.021,[129,129,140,255,131,129,131,131,129],0.019,[133,129,140,210,131,129,131,131,129],[129,129,145,258,131,129,131,131,129],0.037,[133,129,145,208,131,129,131,131,129],[129,129,150,261,131,129,131,131,129],0.06,[133,129,150,159,131,129,131,131,129],[129,129,81,264,131,129,131,131,129],0.061,[133,129,81,161,131,129,131,131,129],[129,129,158,141,131,129,131,131,129],[133,129,158,268,131,129,131,131,129],0.08,[129,129,163,175,131,129,131,131,129],[133,129,163,271,131,129,131,131,129],0.031,[],[230],[],[],[277],"Complete SLAM ATE of sigma-DVO SLAM and PSM SLAM on TUM RGB-D (keyframe counts in the same table not extracted)",{"slug":279,"group":280,"sourceId":5,"sourceLabel":6,"table":281,"selfRows":150,"metrics":282,"seqs":284,"entrants":289,"cells":305,"outcomes":347,"locators":349,"hardware":350,"wordings":351,"notes":352},"psmslam2017-table-2","psmslam2017:Table 2","Table 2",[283],{"label":90,"unit":91,"statistic":92,"alignment":93},[285,286,287,288],{"dataset":99,"sequence":103,"environment":101},{"dataset":99,"sequence":121,"environment":101},{"dataset":99,"sequence":100,"environment":101},{"dataset":99,"sequence":109,"environment":101},[290,292,295,298,301,303,304],{"name":291,"methodId":47,"linkable":66,"proposed":66,"self":66},"RGB-D SLAM [4] (Endres et al.)",{"name":293,"methodId":294,"linkable":70,"proposed":66,"self":66},"Kintinuous [33]","kintinuous2015",{"name":296,"methodId":297,"linkable":70,"proposed":66,"self":66},"MRSMap [28]","mrsmap2014",{"name":299,"methodId":300,"linkable":70,"proposed":66,"self":66},"ElasticFusion [34]","elasticfusion2015",{"name":302,"methodId":47,"linkable":66,"proposed":66,"self":66},"DVO SLAM [14]",{"name":244,"methodId":47,"linkable":66,"proposed":66,"self":66},{"name":246,"methodId":5,"linkable":70,"proposed":70,"self":70},[306,308,309,311,313,314,315,316,318,320,322,324,326,327,328,330,331,333,334,336,337,338,340,341,342,343,344,345],[129,129,129,307,131,129,131,131,129],0.023,[133,129,129,258,131,129,131,131,129],[140,129,129,310,131,129,131,131,129],0.043,[145,129,129,312,131,129,131,131,129],0.02,[150,129,129,253,131,129,131,131,129],[81,129,129,255,131,129,131,131,129],[158,129,129,210,131,129,131,131,129],[129,129,133,317,131,129,131,131,129],0.032,[133,129,133,319,131,129,131,131,129],0.03,[140,129,133,321,131,129,131,131,129],0.042,[145,129,133,323,131,129,131,131,129],0.017,[150,129,133,325,131,129,131,131,129],0.035,[81,129,133,175,131,129,131,131,129],[158,129,133,271,131,129,131,131,129],[129,129,140,329,131,129,131,131,129],0.079,[133,129,140,47,129,129,131,131,129],[140,129,140,332,131,129,131,131,129],0.069,[145,129,140,47,129,129,131,131,129],[150,129,140,335,131,129,131,131,129],0.083,[81,129,140,264,131,129,131,131,129],[158,129,140,161,131,129,131,131,129],[129,129,145,339,131,129,131,131,129],0.084,[133,129,145,47,129,129,131,131,129],[140,129,145,332,131,129,131,131,129],[145,129,145,47,129,129,131,131,129],[150,129,145,194,131,129,131,131,129],[81,129,145,261,131,129,131,131,129],[158,129,145,346,131,129,131,131,129],0.051,[348],"not reported",[281],[],[],[353],"Complete SLAM absolute trajectory error on TUM RGB-D; '-' = no result given",{"slug":355,"group":356,"sourceId":5,"sourceLabel":6,"table":357,"selfRows":150,"metrics":358,"seqs":362,"entrants":373,"cells":380,"outcomes":417,"locators":418,"hardware":419,"wordings":420,"notes":421},"psmslam2017-table-4","psmslam2017:Table 4","Table 4",[359],{"label":360,"unit":91,"statistic":361,"alignment":93},"mean distance from points to nearest ground-truth surface (m)","mean",[363,367,369,371],{"dataset":364,"sequence":365,"environment":366},"ICL-NUIM","lr kt0","synthetic living room",{"dataset":364,"sequence":368,"environment":366},"lr kt1",{"dataset":364,"sequence":370,"environment":366},"lr kt2",{"dataset":364,"sequence":372,"environment":366},"lr kt3",[374,375,376,377,378,379],{"name":291,"methodId":47,"linkable":66,"proposed":66,"self":66},{"name":293,"methodId":294,"linkable":70,"proposed":66,"self":66},{"name":296,"methodId":297,"linkable":70,"proposed":66,"self":66},{"name":302,"methodId":47,"linkable":66,"proposed":66,"self":66},{"name":299,"methodId":300,"linkable":70,"proposed":66,"self":66},{"name":246,"methodId":5,"linkable":70,"proposed":70,"self":70},[381,383,384,385,386,388,390,391,393,395,396,397,399,400,401,403,405,406,408,410,412,414,415,416],[129,129,129,382,131,129,131,131,129],0.044,[133,129,129,250,131,129,131,131,129],[140,129,129,264,131,129,131,131,129],[145,129,129,317,131,129,131,131,129],[150,129,129,387,131,129,131,131,129],0.007,[81,129,129,389,131,129,131,131,129],0.006,[129,129,133,317,131,129,131,131,129],[133,129,133,392,131,129,131,131,129],0.008,[140,129,133,394,131,129,131,131,129],0.14,[145,129,133,264,131,129,131,131,129],[150,129,133,387,131,129,131,131,129],[81,129,133,398,131,129,131,131,129],0.009,[129,129,140,271,131,129,131,131,129],[133,129,140,398,131,129,131,131,129],[140,129,140,402,131,129,131,131,129],0.098,[145,129,140,404,131,129,131,131,129],0.119,[150,129,140,392,131,129,131,131,129],[81,129,140,407,131,129,131,131,129],0.024,[129,129,145,409,131,129,131,131,129],0.167,[133,129,145,411,131,129,131,131,129],0.15,[140,129,145,413,131,129,131,131,129],0.248,[145,129,145,194,131,129,131,131,129],[150,129,145,188,131,129,131,131,129],[81,129,145,398,131,129,131,131,129],[],[357],[],[],[422],"Surface reconstruction accuracy on ICL-NUIM living room with noise: mean distance from reconstructed points to the nearest ground-truth surface (m)",[424],{"group":425,"slug":426,"sourceLabel":6,"table":427,"selfRows":150,"datasets":428},"psmslam2017:Table 5","psmslam2017-table-5","Table 5",[364],1790510664304]