跳到主要內容

臺灣博碩士論文加值系統

(44.192.94.177) 您好!臺灣時間:2024/07/21 18:42
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:陳信宏
研究生(外文):CHEN, HSIN-HUNG
論文名稱:ATopNet:基於卷積神經網路實現視角、光影與速度穩健之視覺地點定位
論文名稱(外文):ATopNet: Pose, Illumination, and Speed-Robust Visual Localization Based on Convolutional Neural Network
指導教授:李志鴻李志鴻引用關係
指導教授(外文):LI, CHIH-HUNG
口試委員:李志鴻張禎元林明璋
口試委員(外文):LI, CHIH-HUNGCHANG, JEN-YUANLIN, MING-CHANG
口試日期:2022-07-01
學位類別:碩士
校院名稱:國立臺北科技大學
系所名稱:機械工程系機電整合碩士班
學門:工程學門
學類:機械工程學類
論文種類:學術論文
論文出版年:2022
畢業學年度:110
語文別:中文
論文頁數:72
中文關鍵詞:視覺地點定位拓撲方法卷積神經網路訓練增強自主移動機器人導航
外文關鍵詞:Visual LocalizationTopological ApproachConvolutional Neural NetworkTraining AugmentationAutonomous Mobile Robot Navigation
相關次數:
  • 被引用被引用:0
  • 點閱點閱:138
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
本文解決了一些在室內環境的關鍵視覺地點定位問題。我們提出了自動拓撲地點定位網路(Automatic Topological Localization Network, ATopNet),適用於使用低分辨率RGB圖像進行視覺地點定位的自主移動機器人(Autonomous Mobile Robot, AMR)。我們的實驗數據證明,ATopNet提供AMR足夠的空間精度與定位準確度進行自主導航,從而允許AMR執行接下來的環境任務。為了有效提升深度卷積神經網路(Convolutional Neural Network, ConvNet)在AMR視覺地點定位的性能,我們研究了提升機器人偏航、環境光線改變、機器人高速行駛與行人遮擋穩健性的自動訓練增強策略。首先推導出了視框與觀察者姿態之間的運動學理論,基於該理論可以自動生成和標註數以萬計以上不同方位的訓練圖像。再來我們提出了利用條件式生成對抗網路將訓練圖像生成陰影效果的方法,也測試和討論了在不同車速的情況下,使用清晰/模糊圖像訓練卷積神經網路對地點定位性能的影響。我們還介紹了使用圖像修復技術來解決行人遮擋問題的方法,以加強視覺地點定位在行人遮擋影響下的識別能力。我們提出了一套拓撲節點與地圖的設計準則,在不增加卷積神經網路訓練負載的前提下,提高AMR在目標位置的定位精度。最後我們展示了ATopNet使用Raspberry Pi 4 + NCS2執行機器人精準地點定位與實時自主導航,並證實本文提出之各項增強策略的效能。
This article addresses some critical visual localization problems in indoor environments. We propose the Automatic Topological Localization Network (ATopNet) for autonomous mobile robots (AMR) using low-resolution RGB images for visual localization. Our experimental results verified that ATopNet provides AMR with sufficient spatial accuracy and localization accuracy for autonomous navigation, allowing AMR to perform subsequent environmental tasks. In order to effectively improve the performance of the deep convolutional neural network (ConvNet) in AMR visual localization, we investigated multitudes of automatic training enhancement strategies to improve the detection robustness subject to robot kinematics, illumination changes, speed effects, and pedestrian occlusion. Firstly, the kinematics theory between the view frame and the observer's pose is derived. Based on this theory, more than tens of thousands of training images with different orientations can be automatically generated and annotated. Next, we proposed using conditional generative adversarial networks to create shadow effects for images utilized in the localization training. We also tested and discussed using clear/blurry images to train convolutional neural networks with an evaluation of the localization performance. We also introduced methods to solve the pedestrian occlusion problem using image inpainting techniques. We further proposed a set of design guidelines for nodes and maps to improve the positioning accuracy of AMR at the target location without increasing the training load of ConvNet. Finally, we exhibited the usage of ATopNet on Raspberry Pi 4 + NCS2 to perform robot precision localization and real-time autonomous navigation, and verified the effectiveness of the enhancement strategies proposed in this paper.
摘 要 i
ABSTRACT ii
誌 謝 iii
目 錄 iv
表目錄 vi
圖目錄 vii
第一章 緒論 1
1.1 前言 1
1.2 研究動機 2
1.3 研究目標 3
1.4 本文貢獻 5
1.5 研究流程與論文架構 5
第二章 文獻探討 7
2.1 視覺定位 7
2.2 神經網路 8
第三章 研究方法 10
3.1 拓撲地點定位 10
3.2 視框運動學 12
3.3 姿態增強 16
3.4 速度變化增強 17
3.5 照明增強 18
3.6 行人干擾增強 19
3.7 嵌入式裝置運算 20
3.8 目標點定位 20
第四章 模型穩健性實驗 22
4.1 實驗環境及硬體 22
4.1.1 地圖 22
4.1.2 機器人 22
4.2 姿態增強 24
4.2.1 視框運動學 24
4.2.2 機器人偏航影響 25
4.2.3 姿態增強策略 26
4.2.4 姿態增強實驗結果 27
4.3 速度變化增強 29
4.3.1 機器人行駛速度影響 29
4.3.2 速度變化增強策略 29
4.3.3 速度變化增強實驗結果 30
4.4 照明增強 30
4.4.1 環境光線影響 30
4.4.2 照明增強策略 32
4.4.3 照明增強實驗結果 34
4.5 行人干擾增強 35
4.5.1 行人遮擋影響 35
4.5.2 行人干擾增強策略 37
4.5.3 行人干擾增強實驗結果 38
4.6 增強功能整合 39
4.7 嵌入式裝置運算加速 42
第五章 拓撲地圖規劃實驗 45
5.1 拓撲節點設計 45
5.2 目標點設計 52
5.3 機器人自主導航展示 58
第六章 結論與未來展望 63
6.1 結論 63
6.2 未來展望 64
參考文獻 66
符號彙編 71

[1]A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet classification with deep convolutional neural networks,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), 2012, pp. 1097–1105.
[2]N. Sunderhauf, S. Shirazi, F. Dayoub, B. Upcroft, and M. Milford, “On the performance of ConvNet features for place recognition,” in Proc. IEEE Int. Conf. Intell. Robots Systs (IROS), 2015, pp. 4297–4304.
[3]Y. Qiao, C. Cappelle, Y. Ruichek, and F. Dornaika, “Visual localization based on sequence matching using ConvNet features,” in Proc. Ann. Conf. IEEE Ind. Electro. Soc. (IECON), 2016, pp. 1067–1074.
[4]A. A. Panchpor, S. Shue, and J. M. Conrad, “A survey of methods for mobile robot localization and mapping in dynamic indoor environments,” in Proc. 2018 Conf. Signal Process. Comm. Eng. Sys. (SPACES), 2018, pp. 138–144.
[5]Y.-F. Hong, Y.-M. Chang, and C. G. Li, “Real-time visual-based localization for mobile robot using structured-view deep learning,” in Proc. IEEE Int. Conf. Auto. Sci. Eng., (CASE), 2019, pp. 1353–1358.
[6]C. G. Li, Y.-F. Hong, P.-K. Hsu, and T. Maneewarn, “Real-time topological localization using structured-view ConvNet with expectation rules and training renewal,” Robot. Auto. Syst., vol. 131, https://doi.org/10.1016/j.robot.2020.103578, 2020.
[7]I. Ulrich and I. Nourbakhsh, “Appearance-based place recognition for topological localization,” in Proc. IEEE Int. Conf. Robot. Auto. (ICRA), 2000, pp. 1023–1029.
[8]A. Kendall, M. Grimes, R. Cipolla, “PoseNet: a convolutional network for real-time 6-DOF camera relocalization,” in Proc. IEEE Int. Conf. Comput. Vision (ICCV), 2015, pp. 2938–2946.
[9]N. Piasco, D. Sidibé, C. Demonceaux, and V. Gouet-Brunet, “A survey on visual-based localization: on the benefit of heterogeneous data,” in Pattern Recognit., vol. 74, pp. 90–109, 2018.
[10]S. Se, D. Lowe, and J. Little, “Mobile robot localization and mapping with uncertainty using scale-invariant visual landmarks,” Int. J. Robotics Research, vol. 21, pp. 735–758, 2002.
[11]H. Bay, T. Tuytelaars, and L. Van Gool, “SURF: speeded up robust features,” in Comput. Vision – ECCV 2006, Lecture Notes Comp. Sci., vol. 3951, Springer, https://doi.org/10.1007/11744023_32.
[12]E. Rublee, V. Rabaud, K. Konolige, and G. Bradski, “ORB: An efficient alternative to SIFT or SURF,” in Proc. IEEE Int. Conf. Comput. Vision, 2011, pp. 2564–2571.
[13]R. Mur-Artal, J. M. M. Montiel, and J. D. Tardos, “ORB-SLAM: A versatile and accurate monocular SLAM system,” IEEE Trans. Robot., vol. 31, no. 5, pp. 1147–1163, 2015.
[14]T. Nicosevici and R. Garcia, “Automatic visual bag-of-words for online robot navigation and mapping,” IEEE Trans. Robot., vol. 28, no. 4, pp. 886–898, 2012.
[15]A. Oliva and A. Torralba, “Modeling the shape of the scene: a holistic representation of the spatial envelope,” Int. J. Comput. Vision, vol. 42, no. 3, pp. 145–175, 2001.
[16]J. Engel, V. Koltun, and D. Cremers, “Direct sparse odometry,” IEEE T. Pattern Anal. Mach. Intell., vol. 40, no. 3, pp. 611–625, 2018.
[17]C. G. Li and Y.-H. Huang, “Deep-trained illumination-robust precision positioning for real-time manipulation of embedded objects,” Int. J. Adv. Manuf. Tech., vol. 111, pp. 2259–2276, 2020.
[18]C. G. Li and Y. M. Chang, “Automated visual positioning and precision placement of a workpiece using deep learning,” Int. J. Adv. Manuf. Tech., vol. 104, no. 9, pp. 4527–4538, 2019.
[19]P.-E. Sarlin, et al., “Back to the feature: learning robust camera localization from pixels to pose,” in Proc. IEEE Conf. Comput. Vision Pattern Recognit. (CVPR), 2021, pp. 3247–3257.
[20]S. Poddar, R. Kottath, and V. Karar, “Evolution of visual odometry techniques,” arXiv.org, 30-Apr-2018. [Online]. Available: https://arxiv.org/abs/1804.11142v1.
[21]D. O. Hebb, The organization of behavior. New York: John Wiley and Sons Inc., 1964.
[22]F. Rosenblatt, “The Perceptron: A probabilistic model for information storage and organization in the brain.,” Psychological Review, vol. 65, no. 6, pp. 386–408, 1958.
[23]K. Fukushima and S. Miyake, “Neocognitron: A self-organizing neural network model for a mechanism of visual pattern recognition,” Competition and Cooperation in Neural Nets, pp. 267–285, 1982.
[24]Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-based learning applied to document recognition,” Proceedings of the IEEE, vol. 86, no. 11, pp. 2278–2324, 1998.
[25]C. Szegedy, Wei Liu, Yangqing Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, and A. Rabinovich, “Going deeper with convolutions,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015.
[26]K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv.org, 19-Dec-2014. [Online].
Available: https://arxiv.org/abs/1409.1556v4.
[27]M. Lin, Q. Chen, and S. Yan, “Network in Network,” arXiv.org, 04-Mar-2014. [Online]. Available: https://arxiv.org/abs/1312.4400.
[28]K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016.
[29]G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, “Densely connected Convolutional Networks,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017.
[30]C. G. Li and L.-P. Zhou, “Training end-to-end steering of a self-balancing mobile robot based on RGB-D image and deep ConvNet,” in Proc. 2020 IEEE/ASME Int. Conf. Adv. Intell. Mechatronics (AIM 2020), 2020, pp. 898–903.
[31]P. Isola, J. Zhu, T. Zhou, and A. A. Efros, “Image-to-image translation with conditional adversarial networks,” in Proc. IEEE Conf. Comput. Vision Pattern Recognit. (CVPR), 2017, pp. 1125–1134.
[32]T. Fan et al., “ Getting robots unfrozen and unlost in dense pedestrian crowds,” IEEE Robot. Auto. Letters, vol. 4, no. 2, pp. 1178–1185, 2019.
[33]Y.-J. Li, H.-H. Chen, and C. G. Li, “Enhancement of visual place recognition for robot localization subject to pedestrian occlusion,” in Proc. IEEE Int. Conf. Auto. Sci. Eng., (CASE), 2021, pp. 348–353.
[34]黎祐任。「行人遮擋對視覺定位之影響及應用生成對抗網路強化定位效能」。碩士論文,國立臺北科技大學機電整合研究所,2022。
[35]M. Bertalmio, G. Sapiro, V. Caselles, and C. Ballester, “Image inpainting,” Proceedings of the 27th annual conference on Computer graphics and interactive techniques - SIGGRAPH '00, 2000.
[36]Y.-C. Hsu, M.-C. Lin, and C. G. Li, “Mobility improvement on the two-wheeled dynamically balanced robot – J4.β,” in Proc. IEEE Int. Conf. Auto. Sci. Eng., (CASE), 2021, pp. 442–447.
[37]林鴻智。「以網宇實體系統實現網路化深度學習與控制的自主移動機器人」。碩士論文,國立臺北科技大學機電整合研究所,2022。
[38]“視角,” Wikipedia, 29-Apr-2022. [Online]. Available: https://zh.m.wikipedia.org/zh-tw/%E8%A6%96%E8%A7%92.

電子全文 電子全文(網際網路公開日期:20270801)
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top