DTAM｜SLAM 點雲圖鑑

DTAM reconstructs dense keyframe depth maps from every pixel with a regularised photometric energy and tracks the monocular camera by whole-image alignment to the dense model on a GPU.

技術屬性

欄位內容為文獻擷取紀錄的原文用語（英文），以原文為據；「未查證」表示本研究尚未讀到該資訊，不代表該方法不具備此能力。

DTAM 的技術屬性
感測輸入	monocular camera
原文測試平台	hand-held Point Grey Flea2 RGB camera in a desktop setting (same setting where PTAM succeeded)
狀態估計	Mapping: per-keyframe inverse depth map minimising a photometric cost volume (average L1 error over tens to hundreds of overlapping frames at S inverse-depth samples) plus an edge-weighted Huber regulariser; the energy is decoupled with an auxiliary variable, solved by primal-dual updates for the convex part and a point-wise exhaustive search over the cost volume whose feasible range shrinks each iteration, with one embedded Newton step for sub-sample accuracy (theta from 0.2 to 1e-4). Tracking: Lucas-Kanade style iterative least squares, first inter-frame rotation on coarse pyramid levels, then 6DOF forward-compositional alignment of the live image to a view synthesised from the dense model, coarse to fine.
資料關聯	direct photometric every-pixel association: cost volume built by projecting reference pixels into overlapping frames for each inverse-depth sample; tracking compares every pixel of the live image with the model-predicted image, rejecting pixels whose photometric error exceeds a threshold that decreases during coarse-to-fine iterations
時間表示	discrete poses
去畸變	不適用
迴圈閉合	none described; the method has no loop detection or map correction, and a relocaliser is mentioned only as disabled during the PTAM comparison
全域最佳化	none; each keyframe's regularised inverse depth map is estimated on its own (no joint optimisation across keyframes is described), and new keyframes are added according to a threshold on the number of pixels in the previous predicted image without visible surface information
地圖表示	overlapping keyframes, each with an RGB reference image, pose, inverse depth map and an M x N x S photometric cost volume; the Fig. 3 example keyframe has nearly 300 x 10^3 estimated points versus about 1000 PTAM point features in the same frame
先驗資訊	pre-calibrated fixed intrinsics with images pre-warped to remove radial distortion; bootstrapped by a standard point-feature stereo method until the first keyframe; static scene and brightness constancy assumed
可輸出幾何	textured dense inverse depth maps; a triangle mesh is computed from each keyframe depth map (oblique edges culled) and used for tracking, forming a surface patchwork with millions of vertices
計算需求	real-time on commodity GPU (abstract); experiments on an NVIDIA GTX 480 (Sec. 3)

使用設備

原文使用的感測器、運算硬體與載具（equipment）。型號保留原文寫法，連結到設備頁中同一型號的歸併名稱；角色依原文用途分為方法輸入、資料集感測器、執行運算平台、參考或真值量測（reference or ground truth）與比較對象設備。

原文使用的設備
類別	型號（原文寫法）	角色	資料集	原文規格	出處
相機	Point Grey Flea2	方法輸入	未標示	30 Hz, 640x480, 24-bit RGB colour, pre-calibrated intrinsics	(Newcombe et al., 2011a, Sec. 3)
運算硬體	NVIDIA GTX 480	執行運算平台	未標示	commodity GPU running mapping and tracking	(Newcombe et al., 2011a, Sec. 3)
運算硬體	i7 quad-core CPU	執行運算平台	未標示	host CPU of the GPU system	(Newcombe et al., 2011a, Sec. 3)

作者報告的優勢與限制

優勢

Dense model reported to give better tracking under rapid motion than a feature-based method (abstract)

限制

Requires GPU hardware (abstract)
Assumes a static scene (abstract)
ORB-SLAM authors note direct dense methods such as DTAM only incrementally expand the map rather than jointly optimising it (Mur-Artal et al., 2015, Sec. IX-B)
Assumes brightness constancy; not robust to real-world global illumination changes (Sec. 3.2 Failure Modes and Future Work)
Initialisation still relies on a feature-based stereo method; a fully dense initialisation is future work (Sec. 2.4)
(observation) Quantitative evaluation is limited to a velocity-profile comparison with PTAM; no ground-truth trajectory or geometric accuracy is reported (Sec. 3.1, Fig. 9)

營建工程相關證據

論文未報告營建或建築量測測試。對本文作為單眼稠密直接法的技術背景，說明稠密表面並不等同已驗證的工程尺度幾何（單眼尺度不可觀）。論文唯一的量化評估是與 PTAM 的線速度曲線比較，未提供地面真值軌跡或幾何精度數值（Sec. 3.1、Fig. 9），因此不能支撐任何工程精度主張；作者並指出系統假設亮度恆定，無法處理真實環境的全域光照變化（Sec. 3.2），因此不適合光照變化大的工地（推論）。

原文驗證環境：受控實驗

報告的性能數據

以下是原文作者報告的性能數值（author-reported results），不是本研究重新量測的結果。每張圖只並列同一個比較組（comparison group，同一張表、同一組實驗設定）內的方法；不同比較組之間的數值不可直接比較，也不構成排名。

本方法共出現在 1 個比較組，合計 1 筆紀錄。

Ghadimzadeh Alamdari et al., 2025 · Table 2 本方法 1 筆

指標Result (run outcome)

資料集與序列Luleå SubT tunnel dataset (Koval et al. 2022)

表格設定（擷取紀錄原文）：Run outcome ('Result' column) of each reviewed vision-based method on the Luleå tunnel test dataset; '+' marks methods not integrated with ROS; the '*' (incompatible with VLP-16) symbol is printed on almost every row (Ghadimzadeh Alamdari et al., 2025, Table 2)

Result (run outcome)，Luleå SubT tunnel dataset (Koval et al. 2022)

只並列這張表在相同設定下報告的方法；以「本方法：」開頭者為本頁方法。失敗、未執行與未報告以標記呈現，不是 0。

失敗
未執行
未報告（沒有數值，不是 0）

這些是 Ghadimzadeh Alamdari et al., 2025 在此表設定下報告的數值（author-reported results），只能在同一個比較組內對照，不代表方法在其他資料或設定下的表現。

統計量：原文未報告；對齊方式：未對齊；單位：無單位；場景：underground tunnel

資料來源作者報告值(Ghadimzadeh Alamdari et al., 2025, Table 2)

數值與出處
方法（原文寫法）	報告值	出處
Mono-SLAM	無數值失敗註記（擷取紀錄）：failed (feature detection and tracking)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
PTAM	無數值失敗註記（擷取紀錄）：failed (initialization for ground floor)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
S-PTAM	無數值未執行註記（擷取紀錄）：未執行 (authors could not run the code)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
OV2SLAM	無數值未執行註記（擷取紀錄）：未執行 (authors could not run the code)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
ORB-SLAM (footnote 1)	無數值未報告註記（擷取紀錄）：success (footnote 1: authors could not run ORB-SLAM 3, so the original ORB-SLAM was used)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
DTAM本方法	無數值未執行註記（擷取紀錄）：未執行 (no publicly available repository)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
LSD-SLAM	無數值失敗註記（擷取紀錄）：failed (feature tracking)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
SVO	無數值失敗註記（擷取紀錄）：failed (tracking)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
DSO	無數值未執行註記（擷取紀錄）：未執行 (authors could not run the code)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
Kinetic Fusion	無數值未執行註記（擷取紀錄）：未執行 (no publicly available repository)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
Dense visual SLAM	無數值失敗註記（擷取紀錄）：failed (feature detection and tracking)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
Elastic Fusion SLAM	無數值未執行註記（擷取紀錄）：未執行 (inconsistent repository)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
Realtime onboard VI estimation	無數值未執行註記（擷取紀錄）：未執行 (no publicly available repository)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
Multi-sensor fusion	無數值未執行註記（擷取紀錄）：未執行 (no publicly available repository)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
SOFT-SLAM	無數值未執行註記（擷取紀錄）：未執行 (no publicly available repository)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
MSCKF	無數值失敗註記（擷取紀錄）：failed (tracking)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
ROVIO	無數值未執行註記（擷取紀錄）：未執行 (authors could not run the code)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
OKVIS	無數值失敗註記（擷取紀錄）：failed (tracking)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
VIORB	無數值未執行註記（擷取紀錄）：未執行 (inconsistent repository)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
S-MSCKF	無數值未執行註記（擷取紀錄）：未執行 (no publicly available repository)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
VINS-Mono	無數值未報告註記（擷取紀錄）：success	(Ghadimzadeh Alamdari et al., 2025, Table 2)
STCM-SLAM	無數值未執行註記（擷取紀錄）：未執行 (no publicly available repository)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
Kimera	無數值失敗註記（擷取紀錄）：failed (feature detection and tracking)	(Ghadimzadeh Alamdari et al., 2025, Table 2)
Yolo-SLAM	無數值未報告註記（擷取紀錄）：other: Result cell reads 'SLAM for dynamic environments'; no run outcome stated	(Ghadimzadeh Alamdari et al., 2025, Table 2)

來源

Newcombe et al., 2011a
Richard A. Newcombe, Steven J. Lovegrove, Andrew J. Davison（2011）DTAM: Dense tracking and mapping in real-time2011 International Conference on Computer Vision (ICCV), pp. 2320-2327
DOI 10.1109/iccv.2011.6126513
同儕審查已出版已讀全文經典查證後修正

回到方法圖鑑