ISC-LOAM (Intensity Scan Context)
強度掃描脈絡（Intensity Scan Context，ISC）是一種同時編碼幾何與 LiDAR 強度的全域描述子：先以距離校正強度，再把 50 m 內的點依方位角與半徑分格，每格保留最大強度，形成一張二維矩陣。檢索分兩階段，先用二值化佔用矩陣做 XOR 比對並估計欄位平移，以處理反向重訪；再以餘弦相似度比對強度結構，最後以時間一致性與 FPFH 加 ICP 的幾何檢查確認迴圈。論文本身只提出迴圈偵測；官方 ISC-LOAM 程式碼再把它接上 LOAM 系列前端與後端最佳化，成為完整的 LiDAR SLAM。
本頁內容
Global LiDAR descriptor that bins calibrated intensity by ring and sector (Intensity Scan Context), retrieved in two stages (binary XOR geometry check with column shift, then intensity cosine similarity) and verified by temporal consistency and FPFH plus ICP; the official ISC-LOAM repository combines it with a LOAM-family front end into a full LiDAR SLAM.
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 with intensity (Velodyne VLP-16 on the warehouse AGV; Velodyne HDL-64E in KITTI)、wheel odometry fused with LiDAR odometry for the front-end trajectory in the warehouse test (Sec. IV-B) |
|---|---|
| 原文測試平台 | wheeled UGV (warehouse AGV)、vehicle (KITTI) |
| 狀態估計 | paper: loop candidates verified by FPFH-based initial alignment followed by ICP (Sec. III-D); the pose-graph back end is not described in the paper. Repository: front end based on LOAM, A-LOAM and F-LOAM and back end based on ISC, with Ceres and GTSAM listed as dependencies (README; LICENSE) |
| 資料關聯 | global descriptor: points within Lmax = 50 m binned into 20 sectors and 60 rings (Table I) with the maximum calibrated intensity per bin; two-stage retrieval: XOR-based binary geometry similarity over column shifts, then column-wise cosine similarity of intensity (Sec. III-B; Sec. III-C) |
| 時間表示 | not described in the paper (descriptor only) |
| 去畸變 | not described in the paper |
| 迴圈閉合 | Intensity Scan Context descriptor with two-stage hierarchical re-identification, temporal consistency over N = 5 neighbouring frames, and FPFH plus ICP geometric verification (Sec. III-C; Sec. III-D; Table I) |
| 全域最佳化 | not described in the paper; the repository lists GTSAM as a dependency, which implies factor-graph optimization (inference from README) |
| 地圖表示 | not described in the paper |
| 先驗資訊 | none; intensity calibration needs a distance-to-intensity mapping collected offline (Sec. III-A) |
| 可輸出幾何 | loop-closure pairs; corrected trajectory and map shown qualitatively for the warehouse test (Fig. 5) |
| 計算需求 | 1.2 ms per query on average; binary geometry matching 0.5 ms on a desktop computer; implemented in C++ with ROS on an Intel NUC mini computer (Sec. I; Sec. III-C; Sec. IV-A; Sec. IV-C). The repository README states 20 Hz for the full SLAM (not verified in the paper) |
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 VLP-16 | 方法輸入 | 未標示 | on the warehouse AGV; intensity rescaled to [0, 1] | (Wang et al., 2020, Sec. III-A; Sec. IV-A) |
| LiDAR | Velodyne HDL-64E | 資料集感測器 | KITTI | on the KITTI car | (Wang et al., 2020, Sec. IV-C) |
| GNSS 接收器 | GPS (KITTI; model not reported) | 參考或真值量測 | KITTI | used to count the total number of loop closures | (Wang et al., 2020, Sec. IV-C) |
| RGB-D 相機 | Intel Realsense r200歸入：Intel RealSense R200 | 比較對象設備 | 未標示 | on the same AGV; the DBoW2 baseline used a front-mounted camera (that this is the R200 is an inference) | (Wang et al., 2020, Sec. IV-A; Sec. IV-B) |
| 輪式或腿式里程計 | AGV wheel odometer (not described) | 方法輸入 | 未標示 | fused with PCL feature-based LiDAR odometry for the front-end trajectory | (Wang et al., 2020, Sec. IV-B) |
| 載具平台 | autonomous guided vehicle for warehouse manipulation | 方法輸入 | 未標示 | maximum speed 1 m/s | (Wang et al., 2020, Sec. IV-A; Fig. 4) |
| 運算硬體 | Intel NUC mini computer (model not reported) | 執行運算平台 | 未標示 | C++ implementation integrated in ROS | (Wang et al., 2020, Sec. IV-A) |
作者報告的優勢與限制
優勢
- On KITTI 00, 02 and 05 its recall is higher than that of every geometry-only LiDAR descriptor compared (Scan Context, GLAROT3D, Cieslewski et al.) and its precision is higher or equal (100% versus 100% for Scan Context on 00 and 05); the text claims higher precision and recall on all three sequences (Table II; Sec. IV-C)
- Handles reverse revisits on KITTI 02 (recall 91% versus 80.6% for DBoW2 and 73% for Scan Context) (Table II)
- 1.2 ms per query on average (Sec. IV-C; Sec. V)
- In the warehouse test it detected a reverse-direction revisit that the DBoW2 visual baseline missed (Sec. IV-B; Fig. 5)
限制
- False positives occur in non-residential stretches with trees on both sides, where geometry and intensity cues are limited (Sec. IV-C)
- Intensity must be calibrated with a distance mapping collected in offline experiments (Sec. III-A)
- Results of the compared LiDAR descriptors were copied from their papers rather than rerun (Sec. IV-C)
- The warehouse experiment is qualitative only (Sec. IV-B; Fig. 5)
營建工程相關證據
論文本身只在倉庫機器人與 KITTI 上測試。施工相關證據來自其他研究：在建築工地的實測評估中，ISC-LOAM 被當成完整 SLAM 基準並報告 APE （Feng et al., 2025）；在基礎設施非破壞檢測的回顧中，作者回報它無法完成軌跡估計（Ghadimzadeh Alamdari et al., 2025）。強度描述子在工地反覆變化的材料與遮蔽下是否穩定，論文沒有驗證。
原文驗證環境：公開基準、受控實驗、已完工建築
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 5 個比較組，合計 32 筆紀錄。以下列出本方法紀錄最多的 4 組，其餘 1 組列在最後，並連到性能比較頁。
Li et al., 2021a · Table II 本方法 13 筆
表格設定（擷取紀錄原文）：KITTI odometry 00-10; mean relative pose error over 100-800 m trajectories (rotation deg/100m / translation %); * marks sequences with loops; LOAM values quoted from its journal paper [19]; other baselines run with open-source code (Li et al., 2021a, Table II)
relative translational error (%)，KITTI odometry · 00*
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 Li et al., 2021a 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(Li et al., 2021a, Table II)
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| LOAM* (from [19]) | 0.78% | (Li et al., 2021a, Table II) |
| FLOAM | 0.92% | (Li et al., 2021a, Table II) |
| ISC-LOAM本方法 | 1.02% | (Li et al., 2021a, Table II) |
| SUMA | 0.77% | (Li et al., 2021a, Table II) |
| SUMA++ | 0.65% | (Li et al., 2021a, Table II) |
| Ours-ODOM原文提出 | 0.59% | (Li et al., 2021a, Table II) |
| Ours-LOOP原文提出 | 0.59% | (Li et al., 2021a, Table II) |
Li et al., 2021a · Table III 本方法 8 筆
指標Absolute Trajectory Error (m)
表格設定（擷取紀錄原文）：KITTI sequences with loops; absolute trajectory error (m); statistic and alignment not stated (Li et al., 2021a, Table III)
Absolute Trajectory Error (m)，KITTI odometry · 00*
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 Li et al., 2021a 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(Li et al., 2021a, Table III)
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| ISC-LOAM本方法 | 1.6 m | (Li et al., 2021a, Table III) |
| SUMA | 1.14 m | (Li et al., 2021a, Table III) |
| SUMA++ | 1.17 m | (Li et al., 2021a, Table III) |
| Ours-ODOM原文提出 | 5.14 m | (Li et al., 2021a, Table III) |
| Ours-LOOP原文提出 | 0.99 m | (Li et al., 2021a, Table III) |
Wang et al., 2020 · Table II 本方法 6 筆
表格設定（擷取紀錄原文）：KITTI sequences 00, 02 (forward and reverse revisits), 05; loop-closure precision and recall (%); Scan Context, GLAROT3D and Cieslewski results copied from their papers, DBoW2 run by the authors; loop ground truth from GPS (Wang et al., 2020, Table II)
Precision (%)，KITTI odometry · sequence 00
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 Wang et al., 2020 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(Wang et al., 2020, Table II)
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| Kim [21] (Scan Context) | 100% | (Wang et al., 2020, Table II) |
| GLAROT3D [17] | 86% | (Wang et al., 2020, Table II) |
| Cieslewski [24] | 92% | (Wang et al., 2020, Table II) |
| Galvez-Lopez [10] (DBoW2) | 100% | (Wang et al., 2020, Table II) |
| Proposed (ISC)本方法原文提出 | 100% | (Wang et al., 2020, Table II) |
Li et al., 2021a · Table IV 本方法 3 筆
指標Absolute Trajectory Error (m)
表格設定（擷取紀錄原文）：Ford Campus Vision and Lidar Dataset, models and parameters tuned on KITTI only; absolute trajectory error (m); statistic and alignment not stated (Li et al., 2021a, Table IV)
Absolute Trajectory Error (m)，Ford Campus Vision and Lidar Dataset · Seq01
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 Li et al., 2021a 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(Li et al., 2021a, Table IV)
| 方法（原文寫法） | 報告值 | 出處 |
|---|---|---|
| FLOAM | 1.61 m | (Li et al., 2021a, Table IV) |
| ISC-LOAM本方法 | 2.3 m | (Li et al., 2021a, Table IV) |
| SUMA | 4.45 m | (Li et al., 2021a, Table IV) |
| SUMA++ | 4.22 m | (Li et al., 2021a, Table IV) |
| Ours-ODOM原文提出 | 1.35 m | (Li et al., 2021a, Table IV) |
| Ours-LOOP原文提出 | 1.35 m | (Li et al., 2021a, Table IV) |
其他比較組
列出其餘 1 個比較組
來源
Wang et al., 2020
（2020）Intensity Scan Context: Coding Intensity and Geometry Relations for Loop Closure Detection2020 IEEE International Conference on Robotics and Automation (ICRA), pp. 2095-2101
DOI 10.1109/icra40945.2020.9196764arXiv 2003.05656程式碼
同儕審查已出版已讀全文近十年查證後修正
相關版本
- 預印本：Intensity Scan Context (arXiv v1) https://arxiv.org/abs/2003.05656
- 程式碼釋出：wh200720041/iscloam: Intensity Scan Context based full SLAM implementation (ISC-LOAM) https://github.com/wh200720041/iscloam
程式碼：https://github.com/wh200720041/iscloam（授權：BSD 3-clause style terms in the repository LICENSE file, which also carries LOAM, A-LOAM and F-LOAM notices (LICENSE read; GitHub license metadata not queried because the API was rate limited)）。有公開程式碼不等於已被重現，也不代表目前版本與論文版本相同。