TLS scanning geometry
作者由簡化的雷達距離方程式推導，指出地面雷射掃描的訊噪比隨入射角餘弦與距離平方下降，並提出入射角係數 cos α 與距離係數；以總體最小平方擬合平面後，把沿雷射束方向的殘差換算為垂直於平面的殘差，藉此分離掃描幾何對單點雜訊的貢獻。實驗先以 Leica HDS6000 掃描 1 m 見方的白色合板（固定 20 m 並旋轉 0° 至 80°，再於 5 m 至 50 m 重複），再以 FARO LS880 HE 從房間中央與角落兩站掃描空房間。房間平均標準差由 3.23 mm 降為去除入射角效應後的 2.55 mm，約 20% 的雜訊來自非零入射角；作者據此建議先以 CAD 圖與初步低解析度掃描評估各測站的入射角與距離，再規劃測站位置。
本頁內容
Models how incidence angle (cos α) and range (squared) degrade TLS signal-to-noise and per-point range noise on planar surfaces, isolates the scan-geometry share with total-least-squares plane fits, and validates it on a rotated reference board (Leica HDS6000, 5 to 50 m, 0° to 80°) and a room scanned from two viewpoints (FARO LS880 HE), where about 20% of the noise stems from non-zero incidence angles (3.23 mm before versus 2.55 mm after correction).
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 | terrestrial laser scanner Leica HDS6000 (reference board experiments)、terrestrial laser scanner FARO LS880 HE (room experiment, 1/4 of full resolution) |
|---|---|
| 原文測試平台 | static terrestrial scanner stations (scanner mounting not described; only the reference board is stated to be on a tripod) |
| 狀態估計 | total least squares plane fitting per segment or per 5°x5° spherical patch; beam-direction residuals converted to orthogonal residuals with the incidence-angle coefficient c_I(α) = cos α and a range coefficient c_R(ρ) derived from a simplified radar range equation under a Lambertian assumption |
| 資料關聯 | 不適用 |
| 時間表示 | 不適用 |
| 去畸變 | 不適用 |
| 迴圈閉合 | 不適用 |
| 全域最佳化 | none |
| 地圖表示 | 不適用 |
| 先驗資訊 | local surface normals from plane fits (or a CAD model) to compute per-point incidence angles; scanner-specific range limits (0 m and 80 m used for the HDS6000) |
| 可輸出幾何 | per-point noise levels (beam direction and orthogonal) and per-patch standard deviation maps shown as net-views; no new point cloud product |
| 計算需求 | offline Matlab on a Dell Precision 390 (dual-core Intel CPU 2.13 GHz, 3 GB RAM, NVIDIA Quadro FX 3500, Windows 7); runtime not reported |
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 |
|---|---|---|---|---|---|
| 地面雷射掃描儀（TLS） | Leica HDS6000 | 方法輸入 | 未標示 | range-model limits d_min 0 m and d_max 80 m used (Sec. 4.2.2); experiments under near-laboratory conditions | (Soudarissanane et al., 2011, Sec. 4, 4.2.2) |
| 地面雷射掃描儀（TLS） | FARO LS880 HE | 方法輸入 | 未標示 | resolution set to 1/4 of full; room point cloud over 20 million points; written 'LS880 HE80' in Sec. 6 | (Soudarissanane et al., 2011, Sec. 5.1, 6) |
| 運算硬體 | Dell Precision 390 | 執行運算平台 | 未標示 | dual-core Intel CPU 2.13 GHz, 3 GB RAM, NVIDIA Quadro FX 3500, Windows 7, Matlab | (Soudarissanane et al., 2011, Sec. 5.2.3) |
| 其他 | 1 x 1 m white coated plywood board on tripod with goniometer | 參考或真值量測 | 未標示 | rotatable horizontally with 2° precision; considered almost Lambertian | (Soudarissanane et al., 2011, Sec. 4, Fig. 4) |
作者報告的優勢與限制
優勢
- Noise growth with incidence angle and range is modelled from the point cloud alone, without external measurements (Sec. 6)
- In the room test about 20% of the measurement noise came from non-zero incidence angles; the average standard deviation fell from 3.23 mm to 2.55 mm after removing the incidence-angle effect (Sec. 5.2.3)
- Supports planning scanner positions by evaluating incidence angles and ranges from a CAD drawing on a first low-resolution scan, for example to meet a precision standard such as 5 mm (Sec. 6)
限制
- Only noise levels are modelled; biases are not studied and angular and range errors are assumed uncorrelated (Sec. 3.2)
- The range model depends on the scanner and only the Leica HDS6000 and FARO LS880 HE were tested (Sec. 6)
- The reference board is not perfectly Lambertian; repetition with Spectralon targets is recommended (Sec. 4.1, 6)
- Planar surfaces are assumed, yet the room walls and floor were not perfectly planar (Sec. 3, 5.2.2)
- Room ranges of 0 to 6.5 m were too short to test the range effect, and near-perpendicular returns may saturate the detector (Sec. 5.2, 5.2.2)
- 9 of 54 board scans at long range and high incidence were unusable (standard deviation above 5 mm or fewer than 100 points) (Sec. 4.2)
營建工程相關證據
作者以 as-built 模型與結構監測作為誤差傳播的應用情境（Sec. 1），並建議先以 CAD 圖與初步低解析度掃描計算入射角與距離，再決定掃描站位以達到指定精度門檻（Sec. 6）。實驗僅在近實驗室條件的板件與空房間進行，未在施工工地驗證。對 SLAM 點雲而言，這提供以入射角與距離作為點位加權或篩選依據的理論基礎（推論）。
原文驗證環境：受控實驗
報告的性能數據
性能數據仍在分批查證，目前尚未收錄此方法的報告值。
來源
Soudarissanane et al., 2011
（2011）Scanning geometry: Influencing factor on the quality of terrestrial laser scanning pointsISPRS Journal of Photogrammetry and Remote Sensing, 66(4), pp. 389-399
DOI 10.1016/j.isprsjprs.2011.01.005
同儕審查已出版已讀全文經典