Scan-vs-BIM object recognition and as-built dimensions
作者改良先前的方法，先以人工選三組以上對應點把工地雷射掃描粗對齊專案 3D CAD 模型，再以新的 ICP 精對齊整個模型，依與各構件表面相符的點數與覆蓋面積判定構件是否被辨識。接著對每個被辨識的構件個別再做 ICP，求得其竣工位姿，並與設計位姿比較，推算柱垂直度與柱間距等尺寸以檢查是否符合容許差。實驗使用加拿大多倫多一座發電廠鋼構廠房施工期間的五次掃描。作者坦承缺乏真值，且位姿偏差與掃描距離相關，結果尚不足以判斷尺寸合規檢查的精度。
本頁內容
Registers site laser scans to the project CAD model with an improved ICP, recognizes objects by matched surface, then refines each object's as-built pose to compute dimensions for tolerance compliance; accuracy could not be established for lack of ground truth.
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 (Trimble GX 3D per Sec. 1.1.2 and ref. [44]) |
|---|---|
| 原文測試平台 | static terrestrial (tripod) |
| 狀態估計 | 不適用 |
| 資料關聯 | closest orthogonal projection of each scan point onto model facets, accelerated with a bounding-volume hierarchy and frustum and back-face culling; ICP-based model fine registration, then per-object fine registration (Sec. 2.2-3.1); point pairs rejected when their distance exceeds tau_D = max(2 sqrt(MSE of previous iteration), 50 mm) or their normals differ by more than 45 deg; iteration stops when the MSE improvement is below 2 mm2 (Sec. 2.2.2) |
| 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 | 不適用 |
| 地圖表示 | point cloud with per-point CAD-object labels |
| 先驗資訊 | 3D CAD model (mesh) of the steel structure, 612 objects (Sec. 2.4) |
| 可輸出幾何 | recognized objects, as-built object poses, derived dimensions such as column plumb and inter-column distances (Sec. 3, Tables 5-6) |
| 計算需求 | CPU implementation, hardware not reported; model fine registration of Scan 4 took about 2 min per iteration over 5 iterations (5 to 10 iterations for the other scans), about 10 min in total; processing Scan 4 including model fine registration, object recognition and as-built pose calculation took about 35 min, 30 min of which for the as-built pose calculation (Sec. 2.4.2, 3.3.2) |
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） | Trimble GX 3D (identified via ref. [44] cited for the scanner used in this research) | 方法輸入 | 未標示 | about 12 mm accuracy at 100 m; maximum horizontal and vertical resolution about 60 microrad (about one point per 3 mm at 50 m); scans 1-4 at 582 microrad and scan 5 at 300 microrad (Table 1) | (Bosché, 2010, Sec. 1.1.2, ref. [44], Table 1) |
作者報告的優勢與限制
優勢
- Model fine registration lowered MSE and increased matched points for all five scans compared with the earlier method (Table 2)
- Object fine registration further lowered MSE (13-37 mm2) (Table 4)
- Recognition is quasi-automated, robust to clutter and occlusion, and efficient (Sec. 4)
- The author states that recall improved for all scans; Table 3 shows higher recall for Scans 2-5 and equal recall for Scan 1 (83%), overall recall 83% vs 80% and precision 93% vs 91%, but lower precision for Scans 4 and 5 (93% vs 94%, 82% vs 84%); gains were small because the manual coarse registrations were already accurate (Sec. 2.4.3, Table 3)
- Model fine registration took about 2 min per iteration on CPU for about 650,000 points and about 20,000 facets; without the acceleration it would take about two orders of magnitude longer (Sec. 2.4.2)
限制
- Assumes each object's shape already complies with tolerances, reasonable only for prefabricated elements (Sec. 3.1)
- No ground truth for as-built poses or dimensions; results not reliable enough to conclude on accuracy for dimensional compliance (Sec. 3.3.1)
- Calculated pose deviations correlate with scanner range (r = 0.45; columns 20-80 m from the scanner), possibly from range-dependent scanner error or fewer recognized points (Sec. 3.3.1)
- Per-object registration is often ill-conditioned because column ends are occluded (Sec. 3.3.1)
- Coarse registration is manual; mesh models lack semantics, so control points were computed manually (Sec. 2.1, 3.3.2)
- Recall and precision rely on object presence identified by manual visual inspection of each scan (footnote to Sec. 2.4.3)
- The expert manual time estimate (a few hours to one day) is not based on field measurements (footnote to Sec. 3.3.2)
- Object poses are refined independently, which may produce clashes between objects (Sec. 5)
營建工程相關證據
施工中鋼構廠房（Portland Energy Center 發電廠專案，多倫多）的現場 TLS 掃描，屬真實工地資料；以 AISC 303-05 與 MNL 135-00 為容許差來源示例（Sec. 3.2）。展示掃描誤差如何傳遞到竣工尺寸，但無獨立參考。
原文驗證環境：施工中工地
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 8 個比較組，合計 129 筆紀錄。以下列出本方法紀錄最多的 4 組，其餘 4 組列在最後，並連到性能比較頁。
Bosché, 2010 · Table 5 本方法 48 筆
表格設定（擷取紀錄原文）：As-built minus as-designed pose of the 16 exterior columns from Scan 4 only; no ground truth or manual survey available; the Delta Z columns (all 0.0 to 0.1 mm) are omitted (Bosché, 2010, Table 5)
Bottom point Delta XYZ (as-built minus as-designed)，PEC steel structure scans · column 1 (estimated range 82.5 m)
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 Bosché, 2010 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 |
|---|---|---|
| New with object fine registration本方法原文提出 | 10.2 mm | (Bosché, 2010, Table 5) |
Bosché, 2010 · Table 6 本方法 44 筆
表格設定（擷取紀錄原文）：Difference between as-built and as-designed distances between structurally connected columns (Scan 4 only); no ground truth (Bosché, 2010, Table 6)
Bottom point Delta XYZ of inter-column distance，PEC steel structure scans · columns 2-1
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 Bosché, 2010 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 |
|---|---|---|
| New with object fine registration本方法原文提出 | -5.2 mm | (Bosché, 2010, Table 6) |
Bosché, 2010 · Table 3 本方法 12 筆
表格設定（擷取紀錄原文）：CAD object recognition (same recognition metric, Surf_min about 0.01 m2 for n = 5) after New vs Old registration; objects present in each scan identified by manual visual inspection (Bosché, 2010, Table 3)
Object recognition recall R%，PEC steel structure scans · Scan 1
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 Bosché, 2010 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(Bosché, 2010, Table 3)
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| New本方法原文提出 | 83% | (Bosché, 2010, Table 3) |
| Old (Bosche et al. [11]) | 83% | (Bosché, 2010, Table 3) |
Bosché, 2010 · Table 2 本方法 10 筆
表格設定（擷取紀錄原文）：Model registration of each site scan to the 3D CAD model (612 objects, 19,478 facets): New = coarse registration plus proposed ICP model fine registration; Old = manual n-point coarse registration only with the earlier matching of Bosche et al. [11] (Bosché, 2010, Table 2)
MSE of matched point-pair distances，PEC steel structure scans · Scan 1
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 Bosché, 2010 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(Bosché, 2010, Table 2)
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| New (model fine registration)本方法原文提出 | 183 mm2 | (Bosché, 2010, Table 2) |
| Old (Bosche et al. [11]) | 637 mm2 | (Bosché, 2010, Table 2) |
其他比較組
來源
Bosché, 2010
（2010）Automated recognition of 3D CAD model objects in laser scans and calculation of as-built dimensions for dimensional compliance control in constructionAdvanced Engineering Informatics, 24(1), 107-118
同儕審查已出版已讀全文經典查證後修正
相關版本
- 預印本：Early version (pre-print) deposited in Heriot-Watt Pure https://pure.hw.ac.uk/ws/files/784859/AEI_2009.pdf