[{"data":1,"prerenderedAt":268},["ShallowReactive",2],{"method-limo2018":3},{"method":4,"reference":56,"equipment":78,"figures":100,"results":101},{"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":25,"sensors":31,"platform":34,"estimator":36,"association":37,"timeModel":38,"deskew":39,"loopClosure":40,"globalOptimization":41,"mapRepresentation":42,"prior":43,"outputGeometry":44,"compute":45,"codeUrl":46,"codeLicense":47,"relatedVersions":48},"limo2018","Graeter et al., 2018","LIMO","LIMO: Lidar-Monocular Visual Odometry",2018,"recent","C07","odometry","LIMO 以單眼相機的特徵追蹤為主，LiDAR 只負責替影像特徵提供深度：先把單次掃描的 LiDAR 點投影到影像，在特徵周圍以深度直方圖切出前景點，再以面積最大的三點平面與視線求交得到特徵深度；地面上的特徵另以 RANSAC 擬合的地面平面處理，超過 30 m 的深度則捨棄。逐影格運動以 PnP 與對極誤差估計後，交給關鍵影格光束法平差，把 LiDAR 深度當成殘差加入，並以關鍵影格與地標篩選、語意剔除動態物件、植生降權及修剪最大殘差維持即時性。系統只做視覺里程計，不含迴圈閉合，也不建立稠密點雲地圖。","Monocular visual odometry that obtains per-feature depth from single LiDAR scans by fitting local planes to foreground-segmented projected points, then refines motion with robustified keyframe bundle adjustment that uses the LiDAR depths as residuals; evaluated on KITTI without loop closure.","full_text_reviewed","peer_reviewed_published","supplementary","論文只在 KITTI 車載道路資料上評估，未涉及施工現場、建築構件或點雲幾何精度；輸出為稀疏地標而非點雲地圖，無法直接作為營建點雲。其參考價值在於說明以 LiDAR 為影像特徵補深度的做法；SDV-LOAM 在同一 KITTI 設定下把 LIMO 列為比較對象 [sdvloam2023]。",[20],"public_benchmark",[22,23,24],"Ranked 13th in translation error and 11th in rotation error on the KITTI odometry benchmark as of 1 March 2018, ahead of ORB-SLAM2 and Stereo LSD-SLAM (Sec. I; Sec. VII)","KITTI evaluation set: LIMO 0.93% translation error and 0.0026 deg\u002Fm rotation error versus 1.22% and 0.0042 deg\u002Fm for the frame-to-frame part Liviodo; bundle adjustment cuts rotation error by nearly 40% (Sec. VI)","Described by the authors as the second best LiDAR-camera method published on KITTI and the best one that does not use ICP-based LiDAR SLAM refinement (Sec. VI)",[26,27,28,29,30],"Highway scenes such as KITTI 01, 12 and 21 give few valid depth estimates because only the road is within usable range, and large optical flow and motion blur hamper tracking (Sec. VI)","Average errors are highest at high speed and at low speed; the low-speed error is attributed to conservative standstill detection (Fig. 7 caption)","The local plane assumption can be violated, giving imprecise depths that the backend must compensate (Fig. 8 caption)","The optimum vegetation weight varies between sequences, which the authors read as a more complex dependence on scene content (Sec. VI)","No loop closure and no dense map output (Sec. I; Fig. 2 caption)",[32,33],"monocular camera (KITTI grayscale images used for feature tracking; camera model not named in the paper)","3D LiDAR (KITTI; model not named in the paper), used only to give depth to image features",[35],"vehicle (KITTI odometry benchmark)","two separate optimizations: (i) frame-to-frame 6-DoF motion from a perspective-n-point cost plus an epipolar cost, each wrapped in a Cauchy loss, used as prior; (ii) windowed keyframe bundle adjustment over reprojection errors, LiDAR depth residuals and a scale regularizer on the oldest motion in the window, with Cauchy losses and a trimmed-least-squares-like removal of the highest residuals after a few iterations (Secs. IV-V, Eq. 7, Algorithm 1)","viso2 feature tracking (about 2000 correspondences in 30-40 ms); feature depth from a local plane through the maximum-area triangle of foreground LiDAR points selected in an image-space neighbourhood by a depth histogram (bin width 0.3 m); ground-plane features use a RANSAC ground fit instead; depth estimates beyond 30 m or at grazing angles rejected; landmarks on dynamic semantic classes rejected and vegetation landmarks weighted (Secs. II, III, V-C, VI)","discrete frames and keyframes; one-shot depth from a single LiDAR scan, without accumulating point clouds over time (Sec. III)","not_reported (no LiDAR motion compensation step is described; KITTI scans are used as provided)","none; the authors state they aim for visual odometry and perform no loop closure (Fig. 2 caption; Sec. I)","none","sparse triangulated landmarks inside the bundle-adjustment window, split into near, middle and far bins and thinned by a voxel filter with median filtering; no dense LiDAR map (Sec. V-C)","LiDAR-camera calibration supplied with KITTI; semantic segmentation network (modified ResNet38) trained on Cityscapes (Sec. VI)","camera poses and a sparse landmark reconstruction; no registered point-cloud map is produced (Fig. 2 caption)","3.5 GHz CPU: Liviodo (frame-to-frame part) at 10 Hz on 2 cores, LIMO at 5 Hz on 4 cores; semantic labels from a modified ResNet38 at 100 ms per image on an Nvidia TitanX Pascal (Sec. VI)","https:\u002F\u002Fgithub.com\u002Fjohannes-graeter\u002Flimo","GPL-3.0 (GitHub license metadata)",[49,53],{"relation":50,"title":51,"doi_or_url":52},"preprint","LIMO: Lidar-Monocular Visual Odometry (arXiv v1, accepted version)","https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.07524",{"relation":54,"title":55,"doi_or_url":46},"code_release","johannes-graeter\u002Flimo",{"id":5,"kind":57,"shortName":7,"title":8,"authors":58,"year":9,"venue":62,"venueType":63,"publisher":64,"volumeIssuePages":65,"doi":66,"arxivId":67,"url":68,"firstPublicDate":69,"publicationStatus":16,"metadataStatus":70,"fulltextStatus":15,"era":10,"classicReason":71,"codeUrl":46,"cluster":11,"topics":72,"mdpi":73,"verification":74,"label":6,"fulltextRoute":75,"versionRead":76,"addedByCensus":77},"method",[59,60,61],"Johannes Graeter","Alexander Wilczynski","Martin Lauer","2018 IEEE\u002FRSJ International Conference on Intelligent Robots and Systems (IROS)","conference","IEEE","pp. 7872-7879","10.1109\u002Firos.2018.8594394","1807.07524","https:\u002F\u002Fdoi.org\u002F10.1109\u002FIROS.2018.8594394","2018-07-19","metadata_verified","not_applicable",[11],false,"confirmed","arXiv","arXiv v1 (2018-07-19, comment 'Accepted at IROS 2018'; only arXiv version); IEEE version of record not read",true,[79,86,90,97],{"category":80,"model":81,"canonical":81,"role":82,"dataset":83,"specs":84,"locator":85},"lidar","KITTI LIDAR (model not named in the paper)","dataset sensor","KITTI odometry","LIDAR point clouds with calibration provided by KITTI; used only for feature depth","Sec. VI",{"category":87,"model":88,"canonical":88,"role":82,"dataset":83,"specs":89,"locator":85},"camera","KITTI camera (model not named in the paper)","grayscale and color images provided by KITTI; grayscale used for tracking",{"category":91,"model":92,"canonical":93,"role":94,"dataset":95,"specs":96,"locator":85},"compute","Nvidia TitanX Pascal","Nvidia Titan-X Pascal","compute for runtime",null,"GPU running the modified ResNet38 semantic segmentation at 100 ms per image",{"category":91,"model":98,"canonical":98,"role":94,"dataset":95,"specs":99,"locator":85},"3.5 GHz CPU (model not reported)","Liviodo on 2 cores at 10 Hz, LIMO on 4 cores at 5 Hz",[],{"totalRows":102,"groupCount":103,"groups":104,"others":267},15,2,[105,182],{"slug":106,"group":107,"sourceId":5,"sourceLabel":6,"table":108,"selfRows":109,"metrics":110,"seqs":132,"entrants":139,"cells":145,"outcomes":171,"locators":172,"hardware":175,"wordings":178,"notes":179},"limo2018-text-sec-vi","limo2018:Text Sec.VI","Text Sec.VI",10,[111,116,119,122,124,126,129],{"label":112,"unit":113,"statistic":114,"alignment":115},"mean translation error (official KITTI metric)","%","mean","not_reported",{"label":117,"unit":118,"statistic":114,"alignment":115},"rotation error (official KITTI metric)","deg\u002Fm",{"label":120,"unit":121,"statistic":115,"alignment":115},"benchmark rank in translation error (as of 1 March 2018)","rank",{"label":123,"unit":121,"statistic":115,"alignment":115},"benchmark rank in rotation error (as of 1 March 2018)",{"label":125,"unit":121,"statistic":115,"alignment":115},"benchmark rank (as of 1 March 2018)",{"label":127,"unit":128,"statistic":115,"alignment":115},"processing rate","Hz",{"label":130,"unit":131,"statistic":115,"alignment":115},"semantic labelling time per image (modified ResNet38)","ms",[133,136],{"dataset":83,"sequence":134,"environment":135},"evaluation (test) set","urban, rural and highway driving",{"dataset":83,"sequence":137,"environment":138},"all","driving",[140,141,143],{"name":7,"methodId":5,"linkable":77,"proposed":77,"self":77},{"name":142,"methodId":5,"linkable":77,"proposed":77,"self":77},"Liviodo (frame-to-frame motion only)",{"name":144,"methodId":5,"linkable":77,"proposed":77,"self":77},"ResNet38 semantic segmentation (component of LIMO)",[146,150,153,155,157,159,162,165,167,168],[147,147,147,148,149,147,149,149,147],0,0.93,-1,[147,151,147,152,149,147,149,149,147],1,0.0026,[151,147,147,154,149,147,149,149,147],1.22,[151,151,147,156,149,147,149,149,147],0.0042,[147,103,147,158,149,151,149,149,147],13,[147,160,147,161,149,103,149,149,147],3,11,[151,163,147,164,149,147,149,149,147],4,30,[151,166,147,109,149,147,147,149,147],5,[147,166,147,166,149,147,151,149,147],[103,169,151,170,149,147,103,149,151],6,100,[],[85,173,174],"Sec. I; Sec. VI; Sec. VII","Sec. I",[176,177,92],"3.5 GHz CPU, 2 cores (model not reported)","3.5 GHz CPU, 4 cores (model not reported)",[],[180,181],"KITTI odometry benchmark evaluation set results as published on the official server (as of 1 March 2018); official KITTI metric; Liviodo is the frame-to-frame part, LIMO the full pipeline with keyframe bundle adjustment","Runtime of the semantic segmentation used for landmark rejection and vegetation weighting",{"slug":183,"group":184,"sourceId":185,"sourceLabel":186,"table":187,"selfRows":166,"metrics":188,"seqs":191,"entrants":203,"cells":215,"outcomes":258,"locators":261,"hardware":263,"wordings":264,"notes":265},"sdvloam2023-table-v","sdvloam2023:Table V","sdvloam2023","Yuan et al., 2023a","Table V",[189],{"label":190,"unit":113,"statistic":114,"alignment":71},"Relative translational error (RTE)",[192,195,197,199,201],{"dataset":83,"sequence":193,"environment":194},"00","urban, highway and country driving",{"dataset":83,"sequence":196,"environment":194},"01",{"dataset":83,"sequence":198,"environment":194},"04",{"dataset":83,"sequence":200,"environment":194},"00-10 average",{"dataset":83,"sequence":202,"environment":194},"11-21 mean (KITTI test set)",[204,207,209,211,213],{"name":205,"methodId":206,"linkable":77,"proposed":73,"self":73},"DEMO","demo2014",{"name":208,"methodId":5,"linkable":77,"proposed":73,"self":77},"LIMO*",{"name":210,"methodId":95,"linkable":73,"proposed":73,"self":73},"Huang et al.",{"name":212,"methodId":95,"linkable":73,"proposed":73,"self":73},"DVL-SLAM",{"name":214,"methodId":185,"linkable":77,"proposed":77,"self":73},"Our VO module",[216,218,220,222,224,226,228,230,232,234,235,237,238,240,242,243,245,247,249,250,252,254,256],[147,147,147,217,149,147,149,149,147],1.05,[147,147,151,219,149,147,149,149,147],1.87,[147,147,103,221,149,147,149,149,147],1.23,[147,147,160,223,149,147,149,149,147],1.16,[147,147,163,225,149,147,149,149,147],1.14,[151,147,147,227,149,147,149,149,147],1.12,[151,147,151,229,149,147,149,149,147],0.91,[151,147,103,231,149,147,149,149,147],0.53,[151,147,160,233,147,147,149,149,147],0.85,[151,147,163,148,149,147,149,149,147],[103,147,147,236,149,147,149,149,147],0.99,[103,147,151,219,149,147,149,149,147],[103,147,103,239,149,147,149,149,147],0.42,[103,147,160,241,149,147,149,149,147],0.94,[160,147,147,148,149,147,149,149,147],[160,147,151,244,149,147,149,149,147],1.47,[160,147,103,246,149,147,149,149,147],0.67,[160,147,160,248,151,147,149,149,147],1.98,[163,147,147,246,149,147,149,149,147],[163,147,151,251,149,147,149,149,147],0.96,[163,147,103,253,149,147,149,149,147],0.77,[163,147,160,255,149,147,149,149,147],0.72,[163,147,163,257,149,147,149,149,147],0.88,[259,260],"average over the sequences reported for LIMO* (00, 01, 04) as printed","as printed; the listed 00-10 values average about 0.98",[262],"Table V; Sec. VII",[],[],[266],"KITTI odometry; relative translational error (%) of LiDAR-assisted depth-enhanced visual odometry; baseline values from the original publications; LIMO* uses semantic information; '-' cells not stored",[],1790510662928]