none｜國立中央大學博碩士論文系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	馮昱翔 Yu-Siang Feng
論文名稱：	Learning-Based Gaussian Belief Propagation for Bundle Adjustment in Visual SLAM
指導教授：	黃志煒 Chi-Wei Huang
口試委員:
學位類別：	碩士 Master
系所名稱：	資訊電機學院 - 通訊工程學系 Department of Communication Engineering
論文出版年：	2022
畢業學年度：	110
語文別：	英文
論文頁數：	43
中文關鍵詞：	視覺SLAM 、光束法平差、高斯置信度傳播、神經網路、空間AI 、機器學習
外文關鍵詞：	Visual SLAM, Bundle adjustment, Gaussian Belief Propagation, Neural network, Spatial AI, Machine learning
相關次數：	點閱：18 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

光束法平差 (bundle adjustment) 是即時優化空間 3D 坐標的主要優化步驟，並且在視覺同步定位與地圖構建 (simultaneously localization and mapping) 中佔了很大一部分的執行時間。儘管目前基於李文柏格-馬夸特 (Levenberg-Marquardt) 的演算法已普遍用於執行快速的光束法平差，但最近採用迭代且在原本硬體執行緩慢的基於高斯置信度傳播的光束法平差展現了其在新興計算平台 intelligence processing unit (IPU) 上快速和準確的潛力。我們提出了一種新穎的架構來使用深度神經網路預測在高斯置信度傳播 (Gaussian belief propagation) 中傳遞的訊息。該模型會提前生成標準高斯置信度中多次迭代後的訊息，以顯著減少所需循環計算的數量。此外，該程序藉由超參數的調整達到收斂，並避免了依賴不固定的阻尼因子來使高斯置信度傳播穩定。與標準高斯置信度傳播相比，基於學習的方法在 GPU 加速下運行速度提高了 17.7 倍，並且有著相同的準確度水平。

Bundle adjustment (BA) is the major optimization step simultaneously refining 3D coordinates and accounts for a large portion of execution time in visual simultaneous localization and mapping (SLAM). While the Levenberg-Marquardt (LM) based algorithms have been commonly used for fast BA, recent solutions adopting iterative and originally slow Gaussian belief propagation (GBP) show its potential to be fast and accurate on emerging computation platforms, Intelligence Processing Unit (IPU). We propose a novel architecture to predict the message passing in GBP with deep neural networks. The model generates messages several iterations ahead to significantly reduce the number of required computation loops. Also, the process converges with hyperparameter tuning and avoids the dependency of an arbitrary damping factor for GBP to be stabilized. Compared with standard GBP, the learning-based approach achieves the same level of accuracy while running 17.7 times faster under GPU acceleration.

Introduction 1
1 Background................................. 1 
2 Motivation.................................. 2 
3 Contribution................................. 3 
4 Framework ................................. 4
Related Works 5
1 Visual Simultaneously Localization and Mapping ...5
2 Graph Neural Network ........................... 6
3 Accelerating Bundle Adjustment...................... 7
4 Gaussian Belief Propagation ........................ 7
Gaussian Belief Propagation Based Bundle Adjustment 10
1 Factor Graph for Bundle Adjustment.................10 
2 Gaussian Belief Propagation Algorithm ...............12
Learning-Based GBP BA 14
1 Bundle Adjustment Architecture with Graph Networks...14
2 Prediction of Gaussian Belief Propagation Message 
 Passing...15
Data Processing and Training Method 18
1 Data Processing............................... 18 
2 Training Method .............................. 18
Experimental Results 21
1 Experimental Setup............................. 21
2 Determine the Number of GN Blocks ................22
3 Comparison with Standard GBP ...................... 22
3.1 Average Reprojection Error Comparison of Testing 
  Segments ... 22
3.2 Average Reprojection Error Comparison in Sequential 
  Keyframes .... 23 
3.3 Computational Time Comparison ................. 24
Conclusion and Future Work 26
1 Conclusion ................................. 26 
2 Future work................................. 26
Bibliography                                27
                                

[1] Andrew J Davison. Futuremapping: The computational structure of spatial ai systems. arXiv preprint arXiv:1803.11288, 2018.
[2] Bill Triggs, Philip F. McLauchlan, Richard I. Hartley, and Andrew W. Fitzgibbon. Bundle adjustment - A modern synthesis. In Bill Triggs, Andrew Zisserman, and Richard Szeliski, editors, Vision Algorithms: Theory and Practice, volume 1883 of Lecture Notes in Computer Science, pages 298–372. Springer, 1999.
[3] Jorge Nocedal and Stephen J Wright. Numerical optimization. Springer, 1999.
[4] Yu Chen, Yisong Chen, and Guoping Wang. Bundle adjustment revisited. arXiv
preprint arXiv:1912.03858, 2019.
[5] Tetsuya Tanaka, Yukihiro Sasagawa, and Takayuki Okatani. Learning To Bundle- Adjust: A Graph Network Approach to Faster Optimization of Bundle Adjustment for Vehicular SLAM. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 6250–6259, October 2021.
[6] Joseph Ortiz, Mark Pupilli, Stefan Leutenegger, and Andrew J Davison. Bundle adjustment on a graph processor. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2416–2425, 2020.
[7] Andrew J Davison and Joseph Ortiz. Futuremapping 2: Gaussian belief propagation for spatial ai. arXiv preprint arXiv:1910.14139, 2019.
[8] Bin Li and Yik-Chung Wu. Convergence analysis of gaussian belief propagation under high-order factorization and asynchronous scheduling. IEEE Transactions on Signal Processing, 67(11):2884–2897, 2019.
[9] Dmitry M Malioutov, Jason K Johnson, and Alan S Willsky. Walk-sums and be- lief propagation in gaussian graphical models. The Journal of Machine Learning Research, 7:2031–2064, 2006.
[10] Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. Layer normalization. arXiv preprint arXiv:1607.06450, 2016.
[11] Garoe Dorta, Sara Vicente, Lourdes Agapito, Neill D.F. Campbell, and Ivor Simp- son. Structured uncertainty prediction networks. Proceedings of the IEEE Confer- ence on Computer Vision and Pattern Recognition, pages 5477–5485, 2018.
[12] Raul Mur-Artal and Juan D. Tardo ́s. ORB-SLAM2: an open-source SLAM system for monocular, stereo, and RGB-D cameras. IEEE Trans. Robotics, 33(5):1255– 1262, 2017.
[13] Andreas Geiger, Philip Lenz, Christoph Stiller, and Raquel Urtasun. Vision meets robotics: The KITTI dataset. Int. J. Robotics Res., 32(11):1231–1237, 2013.
[14] Andrew J Davison, Ian D Reid, Nicholas D Molton, and Olivier Stasse. Monoslam: Real-time single camera slam. IEEE transactions on pattern analysis and machine intelligence, 29(6):1052–1067, 2007.
[15] Javier Civera, Andrew J Davison, and JM Martinez Montiel. Inverse depth parametrization for monocular slam. IEEE transactions on robotics, 24(5):932–945, 2008.
[16] Hauke Strasdat, Jose ́ MM Montiel, and Andrew J Davison. Visual slam: why filter? Image and Vision Computing, 30(2):65–77, 2012.
[17] Georg Klein and David Murray. Parallel tracking and mapping for small ar workspaces. In 2007 6th IEEE and ACM international symposium on mixed and augmented reality, pages 225–234. IEEE, 2007.
[18] Raul Mur-Artal, Jose Maria Martinez Montiel, and Juan D Tardos. Orb-slam: a versatile and accurate monocular slam system. IEEE transactions on robotics, 31(5):1147–1163, 2015.
[19] Marco Gori, Gabriele Monfardini, and Franco Scarselli. A new model for learning in graph domains. In Proceedings. 2005 IEEE international joint conference on neural networks, volume 2, pages 729–734, 2005.
[20] David K Duvenaud, Dougal Maclaurin, Jorge Iparraguirre, Rafael Bombarell, Timothy Hirzel, Ala ́n Aspuru-Guzik, and Ryan P Adams. Convolutional networks on graphs for learning molecular fingerprints. Advances in neural information processing systems, 28, 2015.
[21] Peter Battaglia, Razvan Pascanu, Matthew Lai, Danilo Jimenez Rezende, et al. Interaction networks for learning about objects, relations and physics. Advances in neural information processing systems, 29, 2016.
[22] Peter W Battaglia, Jessica B Hamrick, Victor Bapst, Alvaro Sanchez-Gonzalez, Vinicius Zambaldi, Mateusz Malinowski, Andrea Tacchetti, David Raposo, Adam Santoro, Ryan Faulkner, et al. Relational inductive biases, deep learning, and graph networks. arXiv preprint arXiv:1806.01261, 2018.
[23] Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. Neural message passing for quantum chemistry. In International conference on machine learning, pages 1263–1272. PMLR, 2017.
[24] Xiaolong Wang, Ross Girshick, Abhinav Gupta, and Kaiming He. Non-local neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 7794–7803, 2018.
[25] Siddharth Choudhary, Shubham Gupta, and PJ Narayanan. Practical time bundle adjustment for 3d reconstruction on the gpu. In European Conference on Computer Vision, pages 423–435. Springer, 2010.
[26] Changchang Wu, Sameer Agarwal, Brian Curless, and Steven M Seitz. Multicore bundle adjustment. In CVPR 2011, pages 3057–3064. IEEE, 2011.
[27] Judea Pearl. Probabilistic reasoning in intelligent systems, volume 88. Elsevier, 2014.
[28] Joseph Ortiz, Talfan Evans, and Andrew J. Davison. A visual introduction to gaussian belief propagation. arXiv preprint arXiv:2107.02308, 2021.
[29] Brian C Hall. Lie groups, Lie algebras, and representations: an elementary introduction. Springer, 2015.
[30] Ryan M Eustice, Hanumant Singh, and John J Leonard. Exactly sparse delayed-state filters for view-based slam. IEEE Transactions on Robotics, 22(6):1100–1114, 2006.
[31] Sen Wang, Ronald Clark, Hongkai Wen, and Niki Trigoni. DeepVO: Towards end-to-end visual odometry with deep recurrent convolutional neural networks. In 2017 IEEE International Conference on Robotics and Automation, ICRA 2017, Singapore, Singapore, May 29 - June 3, 2017, pages 2043–2050. IEEE, 2017.
[32] Hugo Larochelle, Dumitru Erhan, Aaron Courville, James Bergstra, and Yoshua Bengio. An empirical evaluation of deep architectures on problems with many factors of variation. Proceedings of the 24th international conference on Machine learn- ing - ICML ’07, pages 473–480, 2007.
[33] Peter J Huber. Robust estimation of a location parameter. In Breakthroughs in statistics, pages 492–518. Springer, 1992.

簡易檢索 / 詳目顯示

相關論文