xGIC 的實驗結果

第五章實驗結果與討論

5.2 視覺輔助慣性測程器

5.2.2 xGIC 的實驗結果

在此部分的實驗，我們分別在室外與室內的環境評估演算法的效能。

 室外的環境

我們以行車紀錄器的固定架將本論文實作的硬體 xGIC 裝設於汽車的擋風玻璃外，

GPS 的天線則以磁鐵將其吸附在車頂上，如 Fig. 5-16 所示，我們以筆記型電腦的 USB 連接 xGIC，作為提供電力與傳輸資料使用，並利用筆記型電腦將 xGIC 所錄到的資料儲存在硬碟上。xGIC 於戶外的拍攝圖如 Fig. 5-17 所示。

Fig. 5-16 本論文實作的硬體 xGIC 裝設於汽車上的實際圖

Fig. 5-17 xGIC 於戶外的拍攝圖

GPS 天線

行車紀錄器 固定架

65 樣以 bucketing[39]的概念挑選固定數量的特徵點來維持一定的計算量。

在說明實驗結果之前，先定義比較演算法效能的方式，我們以 Eulclidean 距離定義位移誤差，並將 RMSE 相對於行進時間以圖的方式呈現，最後以表格比較整個行進過程的 RMSE 和終點的誤差，而在旋轉誤差上，雖然 ADIS 16480 可以直接提供角度的資訊，

但將它裝設在汽車上時，我們發現它所提供的角度資訊是相當不可靠的，這可能是因為磁力計受到嚴重地干擾，因此在這部分並沒有可以作為 ground truth 的角度資訊，也就無法計算旋轉誤差。而與本論文提出的演算法相比的移動軌跡估測包含：(1)GPS，利用內插法將原本 1Hz 的資料填補成 16Hz，並以此作為位置的 ground truth；(2)單純使用慣性量測裝置估測的軌跡(pure IMU navigation)；(3)Geiger 等人於 2011 年所提出的

monocular visual odometry [44]。以下將以汽車環繞交通大學校園一圈評估演算法的效能。

此路徑總長約為 2200 公尺，總計 226 秒，平均速率約為每小時 35 公里。

Fig. 5-18 xGIC 在室外實驗的移動軌跡估測結果

120.995 120.996 120.997 120.998 120.999 121 121.001 24.784 Pure IMU navigation Only gyro+constant velocity Monocular VO(Geiger et al. 2011)

Fig. 5-19 xGIC 在室外實驗的 RMSE 結果比較

Table 5-5 xGIC 在室外實驗的整體 RMSE 和終點誤差結果比較 Algorithm Overall

position RMSE

End point position error Proposed

method 54.5921 m 48.9932 m

Pure

IMU navigation 429.8136 m 1345 m

Monocular

VO [44] 45.1443 m 57.4096 m

觀察 Fig. 5-18 的結果，可以發現 pure IMU navigation 隨著誤差累積，導致結果和 ground truth 相差非常大，但若是不使用加速規積分得到位置，而是只使用陀螺儀並假設汽車是以每秒 10 公尺的速率向前行進，所得的軌跡形狀和 ground truth 較為接近，由此可知 pure IMU navigation 的最主要問題在於無法從加速規得到一個可靠的位移資訊，但陀螺儀的資訊是相對較可靠的。從 Fig. 5-19 和 Table 5-5 則可以得知 proposed method 在本實驗中遜於 monocular visual odometry，主要的原因為在這個 case 下，只使用攝影機或陀螺儀皆可得到可靠的角度資訊，但由於 proposed method 是利用加速規積分估測地圖尺度，而 monocular visual odometry 則是利用已知攝影機裝設於汽車上的高度以及仰角來估測尺度，因此加速規是否包含可靠的位移資訊將會嚴重影響到 proposed method 估測的地圖尺度，不過整體來說，proposed method 在沒有任何關於載具的已知條件下，

所得的軌跡估測結果和 monocular visual odometry 的結果是相近的。若是將 proposed method 的誤差以行進距離取平均，可以得到位移的整體 RMSE 每一公尺為 0.0248 m ，

Position RMSE (m)

Proposed method Pure IMU navigation

Monocular VO(Geiger et al. 2011)

67 Proposed method

Start point [ 0 0 0 ]^T m ground truth 的位置資訊來計算位移誤差。觀察 Fig. 5-20 的結果，可以發現在這個 case 下，proposed method 是勝於 monocular visual odometry，主要原因是因為我們所提出的演算法並沒有限定任何特殊的 motion model，所以 xGIC 是可以在空間上任意運動的，

因此相對於必須在固定高度和仰角的 monocular visual odometry，proposed method 的結果是較好的。proposed method 估測的終點為[0.2082 -0.6224 -1.0983] ^T，實際終點和估測終點間的 Eulclidean 距離約佔整個行進距離的 2.78%，而估測的行進距離約為 47.3566 公尺，與實際行進距離的誤差約佔整個行進距離的 2.95%。 Pure IMU navigation

Monocular VO(Geiger et al. 2011)

第六章研究果與未來展望

6.1 研究成果

本論文提出一套結合單一攝影機與慣性量測裝置的測程器架構，由於使用在三張影像中存在的攝影機幾何限制作為攝影機所提供的量測資訊，使得此方法可以不需要估測特徵點在空間中的位置，也就不需要進行重建環境的演算，並且同時將三張影像分別對應到的攝影機姿態在濾波器中修正，形成一個 multi state constraint Kalman filter，因此是一個在計算量與精確度間取得平衡的 sliding window 式測程法，相較於現存視覺式測程法或是同時定位與地圖建立的方法，本論文提出的架構更符合自我軌跡估測的測程需求並且適用於即時導航系統，而為了有效地排除比對錯誤或是落於移動物體上的特徵點，

以基於三視角幾何的 RANSAC 演算法來挑選 inlier。本論文同時實作了一套整合攝影機、

慣性量測裝置和 GPS 的硬體平台，在這硬體平台中，是以實際電路上的硬體訊號來達成攝影機與慣性量測裝置間的同步，接著本論文實作了攝影機與慣性量測裝置間的空間關係校正演算法並說明其結果。最後以網路上公開的資料集和所實作的硬體平台評估本論文所提出的測程器演算法效能，在資料集部分的實驗結果，以 pure IMU navigation、

monocular visual odometry、stereo visual odometry 和本論文的測程器相比，結果顯示本論文的測程器在地圖上是最接近真實軌跡的，而與 pure IMU navigation、monocular visual odometry 相比，結果顯示經過感測器融合後確實可以得到更可靠的結果，而在本論文實作的硬體平台的實驗結果，以 pure IMU navigation、monocular visual odometry 和本論文的測程器相比，在室外的實驗結果顯示本論文的測程器在沒有任何關於載具的已知條件下，所得的軌跡估測結果和 monocular visual odometry 的結果是相近的，在室內的實驗結果則顯示本論文的測程器即使是在手持的狀況下，所估測的移動軌跡仍有一定的可靠度。

6.2 未來展望

在本論文中，是以特徵點的方式來使用 trifocal tensor，但在 trifocal tensor 原本的推導中，是以線段與線段之間的對應關係推導的，因此未來可以考慮改以線段來使用 trifocal tensor，也就是以線段作為攝影機的量測資訊，線段相較特徵點來說，在許多人造的建築物環境下更為強健，而且在攝影機與慣性量測裝置的感測器融合中，使用線段作為攝影機量測資訊的方式在今年 2013 的 ICRA 才首次被提出[45]，這後續仍有相當多的研究空間。另一方面在測程器的架構中，可以用相對的姿態來作為濾波器的狀態[30]，

這可以使得姿態的不確定性不會被一直傳遞下去，而是被限制在一個大小。

Reference

[1] H. Jwu-Sheng, T. Chin-Yuan, C. Ming-Yuan and S. Kuan-Chun, "IMU-Assisted Monocular Visual Odometry Including the Human Walking Model for Wearable Applications," in Proc. of the IEEE Intl. Conf. on Robotics and Automation (ICRA), Karlsruhe, Germany, May 6-10, 2013.

[2] T. Chin-Yuan, C. Jian-An, and H. Jwu-Sheng, "Unscented Blind Image De-blurring Using Camera with Inertial Measurement Unit," in Proc. of the IEEE Intl. Conf. on Robotics and Biomimetics (ROBIO), Shenzhen, China, Dec. 12-14, 2012, pp. 2096-2101.

[3] K. Jun-Sik, H. Myung, and T. Kanade, "Realtime Afﬁne-photometric KLT Feature Tracker on GPU in CUDA Framework," in Proc. of the IEEE Intl. Conf. on Computer Vision Workshops (ICCV Workshops), Kyoto, Japan, Sep. 27-Oct. 4, 2009, pp. 886-893.

[4] D. Kurz and S. Ben Himane, " Inertial sensor-aligned visual feature descriptors," in Proc.

of the IEEE Intl. Conf. on Computer Vision and Pattern Recognition (CVPR), Colorado Springs, CO, USA, Jun. 20-25, 2011, pp. 161-166.

[5] G. Panahandeh, N. Mohammadiha, and M. Jansson, "Ground Plane Feature Detection in Mobile Vision-Aided Inertial Navigation," in Proc. of the IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems (IROS), Vilamoura, Algarve, Portugal, Oct. 7-12, 2012, pp. 3607-3611.

[6] V. Lippiello, G. Loianno, and B. Siciliano, "MAV Indoor Navigation Based on a Closed-Form Solution for Absolute Scale Velocity Estimation Using Optical Flow and Inertial Data," in Proc. of the IEEE Conf. on Decision and Control and European Control Conference (CDC-ECC), Orlando, FL, USA, Dec. 12-15, 2011, pp. 3566-3571.

[7] L. Kneip, A. Martinelli, S. Weiss, D. Scaramuzza, and R. Siegwart, "Closed-Form Solution for Absolute Scale Velocity Determination Combining Inertial Measurements

and a Single Feature Correspondence," in Proc. of the IEEE Intl. Conf. on Robotics and Automation (ICRA), Shanghai, China, May 9-13, 2011, pp. 4546-4553.

[8] D.-S. Tue-Cuong and A. I. Mourikis, "Estimator Initialization in Vision-aided Inertial Navigation with Unknown Camera-IMU Calibration," in Proc. of the IEEE/RSJ Intl.

Conf. on Intelligent Robots and Systems (IROS), Vilamoura, Algarve, Portugal, Oct. 7-12, 2012, pp. 1064-1071.

[9] T. Lupton and S. Sukkarieh, "Visual-Inertial-Aided Navigation for High-Dynamic

Motion in Built Environments Without Initial Conditions," IEEE Trans. on Robotics, vol.

28, no. 1, pp. 61-76, Feb. 2012.

[10] A. Martinelli, "Vision and IMU Data Fusion: Closed-Form Solutions for Attitude, Speed, Absolute Scale, and Bias Determination," IEEE Trans. on Robotics, vol. 28, no. 1, pp.

44-60, Feb. 2012.

[11] J. B. Kuipers, QUATERNIONS AND ROTATION SEQUENCES. Princeton University Press, 1999.

[12] R. Hartley and A. Zisserman, Multiple View Geometry in computer vision, 2nd ed.

Cambridge University Press, 2008.

[13] J. Sola. (2012, Nov. 6). Quaternion kinematics for the error-state KF [Online].

Available: http://www.joansola.eu/JoanSola/objectes/notes/kinematics.pdf

[14] N. Trawny and S. I. Roumeliotis, "Indirect Kalman Filter for 3D Attitude Estimation,"

University of Minnesota, Dept. of Comp. Sci. & Eng., Tech. Rep. 2005-002, Mar. 2005.

[15] J. Y. Bouguet. (2010, Jul. 9). Camera Calibration Toolbox for Matlab [Online].

Available: http://www.vision.caltech.edu/bouguetj/calib_doc/

[16] Z. Zhengyou, "Flexible Camera Calibration By Viewing a Plane From Unknown Orientations," in Proc. of the IEEE Intl. Conf. on Computer Vision (ICCV), Kerkyra, Corfu, Greece, Sep. 20-25, 1999, pp. 666-673.

[17] J. Heikkila and O. Silven, "A Four-step Camera Calibration Procedure with Implicit Image Correction," in Proc. of the IEEE Computer Society on Computer Vision and Pattern Recognition (CVPR), San Juan, Puerto Rico, Jun. 17-19, 1997, pp. 1106-1112.

[18] F. M. Mirzaei and S. I. Roumeliotis, "A Kalman Filter-Based Algorithm for IMU-Camera Calibration: Observability Analysis and Performance Evaluation," IEEE Trans. on

Robotics, vol. 24, no. 5, pp. 1143-1156, Oct. 2008.

[19] J. Kellyt and G. Sukhatme, "Visual-Inertial Simultaneous Localization, Mapping and Sensor-to-Sensor Self-Calibration," in Proc. of the IEEE Intl. Symp. on Computational Intelligence in Robotics and Automation (CIRA), Daejeon, Korea, Dec. 15-18, 2009, pp.

360-368.

[20] C. D. Hall. (2002). Spacecraft Attitude Dynamics and Control. Lecture Notes, Chapter 3 [Online]. Available: http://www.dept.aoe.vt.edu/~cdhall/courses/aoe4140/

[21] P. Corke, J. Lobo, and J. Dias, "An Introduction to Inertial and Visual Sensing," Intl.

Journal of Robotics Research, vol. 26, no. 6, pp. 519-535, Jun. 2007.

[22] G. Klein and D. Murray, "Parallel Tracking and Mapping for Small AR Workspaces," in Proc. of the IEEE/ACM Intl. Symp. on Mixed and Augmented Reality (ISMAR), Nara, Japan, Nov. 13-16, 2007, pp. 225-234.

[23] G. Nützi, S. Weiss, D. Scaramuzza, and R. Siegwart, "Fusion of IMU and Vision for Absolute Scale Estimation in Monocular SLAM," Journal of Intelligent & Robotic Systems, vol. 61, no. 1-4, pp. 287-299, Jan. 2011.

[24] S. Weiss and R. Siegwart, "Real-Time Metric State Estimation for Modular

Vision-Inertial Systems," in Proc. of the IEEE Intl. Conf. on Robotics and Automation (ICRA), Shanghai, China, May 9-13, 2011, pp. 4531-4537.

[25] L. Kneip, S. Weiss, and R. Siegwart, "Deterministic Initialization of Metric State

Estimation Filters for Loosely-Coupled Monocular Vision-Inertial Systems," in Proc. of

the IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems (IROS), San Francisco, CA, USA, Sep. 25-30, 2011, pp. 2235-2241.

[26] D. Scaramuzza and F. Fraundorfer, "Visual Odometry [Tutorial]," IEEE Robotics Automation Magazine, vol. 18, no. 4, pp. 80-92, Dec. 2011

[27] D. Nister, O. Naroditsky, and J. Bergen, "Visual Odometry," in Proc. of the IEEE Computer Society Conf. on Computer Vision and Pattern Recognition (CVPR), vol. 1, Washington, DC, USA, Jun. 27-Jul. 2, 2004, pp. 652-659.

[28] A. J. Davison, I. D. Reid, N. D. Molton, and O. Stasse, "MonoSLAM: Real-Time Single Camera SLAM," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 29, no.

6, pp. 1052-1067, Jun. 2007.

[29] J. Civera, A. J. Davison, and J. Montiel, "Inverse Depth Parametrization for Monocular SLAM," IEEE Trans. on Robotics, vol. 24, no.5, pp. 932-945, Oct. 2008.

[30] J. Civera, O. G. Grasa, A. J. Davison, and J. M. M. Montiel, "1-Point RANSAC for EKF Filtering. Application to Real-Time Structure from Motion and Visual Odometry,"

Journal of Field Robotics, vol. 27, no. 5, pp. 609-631, Sep. 2010.

[31] P. Pinies, T. Lupton, S. Sukkarieh, and J. D. Tardos, "Inertial Aiding of Inverse Depth SLAM using a Monocular Camera," in Proc. of the IEEE Intl. Conf. on Robotics and Automation (ICRA), Roma, Italy, Apr. 10-14, 2007, pp. 2797-2802.

[32] A. I. Mourikis and S. I. Roumeliotis, "A Multi-State Constraint Kalman Filter for Vision-aided Inertial Navigation," in Proc. of the IEEE Intl. Conf. on Robotics and Automation (ICRA), Roma, Italy, Apr. 10-14, 2007, pp. 3565-3572.

[33] L. Mingyang and A. I. Mourikis, "Improving the Accuracy of EKF-Based Visual-Inertial Odometry," in Proc. of the IEEE Intl. Conf. on Robotics and Automation (ICRA), St. Paul, Minnesota, USA, May 14-18, 2012, pp. 828-835.

[34] L. Mingyang and A. I. Mourikis, "Vision-aided Inertial Navigation for Resource -constrained Systems," in Proc. of the IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems (IROS), Vilamoura, Algarve, Portugal, Oct. 7-12, 2012, pp. 1057-1063.

[35] Y. Ying-Kin, W. Kin Hong, M. M. Y. Chang, and O. Siu Hang, "Recursive Camera Motion Estimation with Trifocal Tensor," IEEE Trans. on Systems Man and Cybernetics Part B Cybernetics, vol. 36, no. 5, pp. 1081-1090, Oct. 2006.

[36] D. D. Diel, P. DeBitetto, and S. Teller, "Epipolar Constraints for Vision-Aided Inertial Navigation," in Proc. of the IEEE Workshops on Application of Computer Vision (WACV/MOTIONS), vol. 2, Breckenridge, CO, USA, Jan. 5-7, 2005, pp. 221-228.

[37] D. Zachariah and M. Jansson, "Camera-aided inertial navigation using epipolar points,"

in Proc. of the IEEE/ION Position Location and Navigation Symposium (PLANS), Indian Wells, CA, USA, May 4-6, 2010, pp. 303-309.

[38] J. O. Nilsson, D. Zachariah, M. Jansson, and P. Handel, "Realtime implementation of visual-aided inertial navigation using epipolar constraints," in Proc. of the IEEE/ION Position Location and Navigation Symposium (PLANS), Myrtle Beach, SC, USA, Apr.

23-26, 2012, pp. 711-718.

[39] B. Kitt, A. Geiger, and H. Lategahn, "Visual Odometry based on Stereo Image Sequences with RANSAC-based Outlier Rejection Scheme," in Proc. of the IEEE Intelligent

Vehicles Symposium (IV), La Jolla, CA, USA, Jun. 21-24, 2010, pp. 486-492.

[40] E. Asadi and C. L. Bottasso, "Tightly-coupled vision-aided inertial navigation via trifocal constraints," in Proc. of the IEEE Intl. Conf. on Robotics and Biomimetics (ROBIO), Shenzhen, China, Dec. 12-14, 2012, pp. 85-90.

[41] V. Indelman, P. Gurfil, E. Rivlin, and H. Rotstein, "Real-Time Vision-Aided Localization and Navigation Based on Three-View Geometry," IEEE Trans. on Aerospace and

在文檔中使用多個狀態限制卡爾曼濾波器與三焦張量幾何之視覺輔助慣性測程器 (頁 74-0)

第五章 實驗結果與討論