臺灣博碩士論文加值系統

目前大多數的地板區域偵測演算法仍然依賴著深度或一些外加的感測器，相對的需要消耗更多的計算資源並且增加無人地面載具的總重量。在本論文中，提出了一個創新的架構，讓無人地面載具能夠僅使用單張影像輸入來判斷出地板區域。由於在室內環境中有花紋地板、陰影和反光，使得區分地板區域更加困難，因此我們提出的演算法結合了全卷積神經網路(FCN)、Canny邊緣偵測以及CCL等值標號來辨識出地板區域，並使用NVIDIA Jetson TX2平台來實現硬體加速使得演算法達到即時辨識的效能。在實驗結果中，我們使用public MIT Scene dataset和各種室內環境的數據集來驗證我們提出演算法的正確率，而我們提出的演算法在沒有外加任何感測器情況下，正確率平均可達到97.1%。

In this paper, a new algorithm combines a Full Convolutional Network (FCN) to identify a rough floor area, and then using Canny Edge Detection to identify a rough floor area, and intersect the two results. In order to achieve real-time performance, the proposed algorithm has been implemented on an NVIDIA Jetson TX2 embedded platform. In experimental results, a public MIT scene dataset and indoor database were used to verify detection accuracy. Compared to recent works, the proposed FCN based algorithm accuracy can reach up to 97.1% on average without the assistance of any other physical sensors, such as RGB-D or laser ranger sensors.

摘要..................................i
Abstract..................................ii
誌謝..................................iii
目錄..................................iv
表目錄..................................v
圖目錄..................................vi
第一章 緒論..................................1
1.1 研究背景與動機..................................1
1.2 研究目的與方法..................................1
1.3 文獻回顧與探討..................................1
第二章 地板區域辨識演算法..................................8
2.1 表面紋理特徵提取演算法..................................9
2.1.1 超像素演算法..................................9
2.2 圖片特定幾何區域..................................12
2.2.1 Canny邊緣偵測..................................13
2.2.2 物件標籤化..................................15
2.3 全卷積神經網路..................................18
2.4 特定表面地板區域..................................26
第三章 演算法實驗結果..................................27
第四章 嵌入式系統應用與機器人載具實現..................................33
4.1 ROS開發與實現..................................35
第五章 結論與未來展望..................................38
參考文獻..................................39
Extended Abstract..................................41

[1]L. Wang, Z. Zhou, J. Wu, Y. Liu, and X. Zhao, “Support-plane estimation for floor detection to understand regions’ spatial organization,” in IEEE International Conference on Robotics and Biomimetics, Dec 2014, pp. 2576–2581.
[2]M. Seki and Y. Sugaya, “Floor-wall boundary detection from projected ellipses for autonomous robot navigation,” in International Conference On Advanced Informatics: Concepts, Theory And Application, Aug 2016, pp. 1–4.
[3]C. Chun, D. Park, W. Kim, and C. Kim, “Floor detection based depth estimation from a single indoor scene,” in IEEE International Conference on Image Processing, Sept 2013, pp. 3358–3362.
[4]S. Kumar, M. S. Karthik, and K. M. Krishna, “Markov random field based small obstacle discovery over images,” in IEEE International Conference on Robotics and Automation, May 2014, pp. 494–500.
[5]Y. Li and S. T. Birchfield, “Image-based segmentation of indoor corridor floors for a mobile robot,” in IEEE/RSJ International Conference on Intelligent Robots and Systems, Oct 2010, pp. 837–843.
[6]V. Hedau, D. Hoiem, and D. Forsyth, “Recovering the spatial layout of cluttered rooms,” in IEEE International Conference on Computer Vision, Sept 2009, pp. 1849–1856.
[7]S. Aggarwal, A. M. Namboodiri, and C. V. Jawahar, “Estimating floor regions in cluttered indoor scenes from first person camera view,” in International Conference on Pattern Recognition, Aug 2014, pp. 4275–4280.
[8]J. Long, E. Shelhamer, & Darrell, T. 2015. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3431-3440.
[9]N. Yang and C.C. Sun ”A New Fast Estimating Floor Region based on Image Segmentation for Smart Rovers” VLSI CAD 2017, Taiwan.
[10]L. C. Chen, G. Papandreou, I. Kokkinos, K. Murphy and A. L. Yuille, (2018). Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs. IEEE transactions on pattern analysis and machine intelligence, 40(4), 834-848.
[11]R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua, and S. Sstrunk, “Slic superpixels compared to state-of-the-art superpixel methods,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 34, no. 11, pp. 2274–2282, Nov 2012.
[12]D. Hoiem, A. A. Efros, and M. Hebert, “Recovering surface layout from an image,” International Journal of Computer Vision, vol. 75, no. 1, pp. 151–172, 2007.
[13]J. Canny, “A computational approach to edge detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PAMI-8, no. 6, pp. 679–698, Nov 1986.
[14]A. AbuBaker, R. Qahwaji, S. Ipson, and M. Saleh, “One scan connected component labeling technique,” in IEEE International Conference on Signal Processing and Communications, Nov 2007, pp. 1283–1286.
[15]A. Quattoni and A. Torralba, “Recognizing indoor scenes,” in IEEE Conference on Computer Vision and Pattern Recognition, June 2009, pp. 413–420.
[16]M. Kubat and S. Matwin, “Addressing the curse of imbalanced training sets: one-sided selection,” in International Conference on Machine Learning. Morgan Kaufmann, 1997, pp. 179–186.
[17]J. M. O’Kane, A Gentle Introduction to ROS. Independently published, Oct. 2013, available at http://¬www.cse.sc.edu/¬~jokane/¬agitr/¬.
[18]B. Gerkey, W. Smart, and M. Quigley, Programming Robots with ROS: A Practical Introduction to the Robot Operating System. O’Reilly Media, 2015.