HDL-64E S2 calibration
作者以平面特徵約束的 Gauss-Helmert 最小二乘法，在車載動態資料中同時估計 Velodyne HDL-64E S2 的視準角、槓桿臂與每顆雷射的內部校正參數。資料取自 2010 年在德州 The Woodlands 停車場以原型車載系統（IMAR AirSurv-RQH IMU、Novatel OEM-4 GPS）多次通過所得，萃取 75 個約 4 m² 的平面。可觀測性分析顯示各雷射水平旋轉修正與航向視準角近乎完全相關、z 向槓桿臂幾乎不可觀測、水平與垂直偏移也高度相關，因此移出估計。縮減參數後，平面閉合差 RMSE 由僅做視準校正的 0.047 m 降至 0.034 m（加入靜態校正的水平旋轉修正後為 0.030 m）；限 25 m 內時由 0.037 m 降至 0.025 m，接近 Riegl LMS-Q120i 的 0.020 m，但不及 VZ-400 的 0.013 m。
本頁內容
In-situ kinematic calibration of the HDL-64E S2 by planar-constrained Gauss-Helmert adjustment of boresight, lever arm and per-laser parameters; several parameters prove unobservable, and the reduced model cuts planar misclosure RMSE by about 30% (0.047 to 0.034 m; 0.025 m within 25 m).
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。
| 感測輸入 | 3D LiDAR (Velodyne HDL-64E S2)、IMU (IMAR AirSurv-RQH, navigation grade)、GNSS (Novatel OEM-4 dual-frequency GPS receiver) |
|---|---|
| 原文測試平台 | vehicle-mounted prototype mobile scanning platform (Fig. 2), multiple passes through an open parking lot in The Woodlands, Texas, on 23 March 2010 |
| 狀態估計 | planar-feature-conditioned combined (Gauss-Helmert) least-squares adjustment estimating boresight angles, lever arm and per-laser interior parameters together with plane coefficients; after observability analysis the per-laser horizontal rotation corrections, horizontal and vertical offsets and the z lever arm were removed, leaving roll, pitch, yaw, horizontal lever arm and three parameters per laser |
| 資料關聯 | semi-automatic extraction of 75 planar surfaces of about 4 m2 each (mostly horizontal or vertical, about 15 near 45 deg), thinned to about 750 returns per plane at 5 to 100 m range |
| 時間表示 | 不適用 |
| 去畸變 | implicit: each return is georeferenced with the GNSS/INS pose at its own measurement time (Eq. 1); no separate deskew step |
| 迴圈閉合 | 不適用 |
| 全域最佳化 | none |
| 地圖表示 | 不適用 |
| 先驗資訊 | smoothed best estimate of trajectory from tightly coupled GPS/INS processing (forward and reverse separation below 2 cm, at least 8 satellites); Velodyne factory calibration as initial model; factory horizontal and vertical offsets held fixed; in one variant horizontal rotation corrections taken from a static calibration |
| 可輸出幾何 | kinematic interior calibration and boresight parameters with their precision, and planar misclosure statistics of the georeferenced point cloud |
| 計算需求 | 不適用 |
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。
| 類別 | 型號（原文寫法） | 角色 | 資料集 | 原文規格 | 出處 |
|---|---|---|---|---|---|
| LiDAR | Velodyne HDL-64E S2歸入：Velodyne HDL-64E | 方法輸入 | 未標示 | 64 lasers; 360 deg by 26.8 deg FOV; range 50 m (10% reflectivity) and 120 m (80%); 1.5 cm range precision (1 sigma); 0.09 deg azimuth resolution; 1,333,333 measurements per second; 905 nm, 5 ns pulse, 2.0 mrad beam divergence; head up to 900 rpm (Table 6 lists 20 mm range accuracy) | (Glennie, 2012, Introduction; Table 1; Table 6) |
| LiDAR | Riegl LMS-Q120i | 比較對象設備 | 未標示 | range accuracy 15 mm, beam divergence 2.7 mrad, angular resolution 0.01 deg, 10,000 Hz; same platform and trajectory, boresight-only adjustment | (Glennie, 2012, Table 6; Table 7) |
| LiDAR | Riegl VZ-400 | 比較對象設備 | 未標示 | range accuracy 5 mm, beam divergence 0.3 mrad, angular resolution 0.0005 deg, 125,000 Hz; same platform and trajectory, boresight-only adjustment | (Glennie, 2012, Table 6; Table 7) |
| 慣性量測單元（IMU） | IMAR AirSurv-RQH | 方法輸入 | 未標示 | navigation grade; expected attitude noise about 0.005 deg roll and pitch, 0.01 deg yaw | (Glennie, 2012, Experimental Description; Accuracy of Parameter Estimates) |
| GNSS 接收器 | Novatel OEM-4 | 方法輸入 | 未標示 | dual-frequency GPS receiver; at least eight satellites tracked; tightly coupled forward and reverse processing with separation below 2 cm | (Glennie, 2012, Experimental Description) |
| 載具平台 | prototype mobile scanning platform (vehicle; model not reported) | 方法輸入 | 未標示 | carries the HDL-64E S2, IMU and GPS receiver with raw data logging | (Glennie, 2012, Experimental Description; Fig. 2) |
作者報告的優勢與限制
優勢
- Full kinematic calibration lowered overall planar misclosure RMSE from 0.047 m (boresight only) to 0.034 m, and to 0.030 m with the static horizontal rotation correction (Table 4)
- Within 25 m range, RMSE fell from 0.037 m to 0.025 m (0.023 m with static correction), close to the Riegl LMS-Q120i at 0.020 m (Tables 5, 7)
- Removing returns with incidence angle above 70 deg lowered RMSE from 0.030 m to 0.022 m (Comparison section)
- Variance factor 0.69; roll and yaw estimated better than IMU attitude noise, distance offset to about a quarter of the ranging noise (Table 3)
限制
- Per-laser horizontal rotation corrections are almost perfectly correlated with the heading boresight (above 0.9) and cannot be estimated kinematically (Observability section)
- The z lever arm is very weakly observable with only four plane orientations (Observability section)
- Horizontal and vertical offsets correlate about 0.9 with the rotation corrections and are held at factory values (Observability section)
- Systematic residuals remain, mostly at high incidence angles; a second-order range term may be needed and range offset may drift with temperature (Comparison; Summary)
- Single calibration dataset from one site and day (inference from the Experimental Description; not stated as a limitation by the author)
營建工程相關證據
原文未報告
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 4 個比較組，合計 15 筆紀錄。
Glennie, 2012 · Table 4 本方法 6 筆
資料集與序列authors' parking-lot calibration dataset (23 March 2010) · all ranges
表格設定（擷取紀錄原文）：Planar misclosure over 75 planes (about 750 returns each, 5 to 100 m range) from one kinematic dataset; columns: factory interior calibration with global boresight only, full kinematic calibration, kinematic calibration with static horizontal correction (Glennie, 2012, Table 4)
RMSE planar misclosure，authors' parking-lot calibration dataset (23 March 2010) · all ranges
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。
按 Tab 進入圖表後，用上下方向鍵逐一瀏覽各類別，Esc 關閉提示框；也可開啟表格檢視閱讀全部數值。
這些是 Glennie, 2012 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。
資料來源作者報告值(Glennie, 2012, Table 4)
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| Global boresight only | 0.047 m | (Glennie, 2012, Table 4) |
| Kinematic Calibration本方法原文提出 | 0.034 m | (Glennie, 2012, Table 4) |
| Kinematic Calibration with H_o^i from Static Analysis (text: horizontal rotation correction from static calibration)本方法原文提出 | 0.03 m | (Glennie, 2012, Table 4) |
Glennie, 2012 · Table 5 本方法 6 筆
資料集與序列authors' parking-lot calibration dataset (23 March 2010) · range limited to 25 m
表格設定（擷取紀錄原文）：Same planar misclosure statistics as Table 4 but limited to points within 25 m of the scanner (Glennie, 2012, Table 5)
RMSE planar misclosure，authors' parking-lot calibration dataset (23 March 2010) · range limited to 25 m
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。
按 Tab 進入圖表後，用上下方向鍵逐一瀏覽各類別，Esc 關閉提示框；也可開啟表格檢視閱讀全部數值。
這些是 Glennie, 2012 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。
資料來源作者報告值(Glennie, 2012, Table 5)
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| Global boresight only | 0.037 m | (Glennie, 2012, Table 5) |
| Kinematic Calibration本方法原文提出 | 0.025 m | (Glennie, 2012, Table 5) |
| Kinematic Calibration with H_o^i from Static Analysis (text: horizontal rotation correction from static calibration)本方法原文提出 | 0.023 m | (Glennie, 2012, Table 5) |
Glennie, 2012 · Table 7 本方法 2 筆
指標RMSE
資料集與序列authors' parking-lot calibration datasets · range limited to 25 m
表格設定（擷取紀錄原文）：Planar RMSE residuals for points within 25 m; Riegl datasets collected at other times on the same platform, trajectory and similar GPS constellation, with global boresight adjustment only (Glennie, 2012, Table 7)
RMSE，authors' parking-lot calibration datasets · range limited to 25 m
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。
按 Tab 進入圖表後，用上下方向鍵逐一瀏覽各類別，Esc 關閉提示框；也可開啟表格檢視閱讀全部數值。
這些是 Glennie, 2012 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。
資料來源作者報告值(Glennie, 2012, Table 7)
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| Velodyne Kinematic Calibration本方法原文提出 | 0.025 m | (Glennie, 2012, Table 7) |
| Velodyne Kinematic Calibration H From Static Calibration本方法原文提出 | 0.023 m | (Glennie, 2012, Table 7) |
| Riegl LMS-Q120i (boresight only) | 0.02 m | (Glennie, 2012, Table 7) |
| Riegl VZ-400 (boresight only) | 0.013 m | (Glennie, 2012, Table 7) |
Glennie, 2012 · Text Comparison section 本方法 1 筆
指標overall RMS planar misclosure
資料集與序列authors' parking-lot calibration dataset (23 March 2010) · all ranges, incidence angle 70 deg or less
表格設定（擷取紀錄原文）：Effect of removing returns with incidence angle above 70 deg on the overall planar misclosure (all ranges, full calibration with static correction) (Glennie, 2012, Text Comparison section)
overall RMS planar misclosure，authors' parking-lot calibration dataset (23 March 2010) · all ranges, incidence angle 70 deg or less
這張表在此指標與資料序列只列出本方法一筆，沒有可並列的其他方法，因此不畫圖，數值與出處見下表。這是 Glennie, 2012 在此表設定下報告的數值（author-reported results），不代表方法在其他資料或設定下的表現。
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| Kinematic calibration with static horizontal correction, high-incidence points removed本方法原文提出 | 0.022 m有附註註記（擷取紀錄）：reduced from 0.030 m (almost 30%) | (Glennie, 2012, Comparison of Velodyne with Other Mobile Scanners) |
來源
Glennie, 2012
（2012）Calibration and Kinematic Analysis of the Velodyne HDL-64E S2 Lidar SensorPhotogrammetric Engineering & Remote Sensing, 78(4), pp. 339-347
同儕審查已出版已讀全文經典