基於視覺定位與環境感知之盲人導航系統

Vision-based Global Localization and Scene Understanding for Navigation for the Visually Impaired

杜品蓉(Ping-Jung Duh)

指導教授 ： 陳冠文

國立交通大學/資訊學院/多媒體工程研究所/碩士(2018年)

摘要

平常我們走在路上能輕易的看到環境並決定要如何行走，對於視障者而言，要得 知這些環境的資訊卻不是這麼容易;近年來，幫助視障者在陌生環境下導航的相 關研究逐漸增多，但是還未出現能實際了解他們需要的資訊並且透過純視覺的方 式使用一般相機作為感測器的系統。為了要達到這樣的系統功能，我們首先進行 了訪談再加以統整作為設計系統的依據，透過視覺定位的方式加上深度學習的系 統的搭配，系統能讓視障者除了知道目前的位置和方向，還能即時得知當下環境 的訊息，包含有哪些該注意東西;在透過易用性測試實驗後，也得到在陌生環境 下本系統有一定程度的準確性的回饋。本研究利用新的方式呈現導航系統，提供 未來在相關領域的研究能有新的想法概念。

關鍵字

盲人導航 ； 視覺定位 ； 環境感知

並列摘要

Recognizing the surroundings can be a challenge when one is incapable of sight. In the past few years, guidance for the visually impaired has received considerable attention. Several studies have proposed methods to help visually impaired people overcome the difficulty in traveling to unfamiliar places. However, most of them do not provide the information that the visually impaired really need. To date, there have not been many complete systems developed for navigating both indoors and outdoors effectively with a monocular camera, to the best of our knowledge. In this study, we firstly interview the visually impaired to understand their needs, with regard to the system design. To fulfill these requirements, we propose a vision-based approach for establishing the navigation system, in which a novel global localization method is proposed and is combined with image segmentation techniques for better scene understanding. As a result, the visually impaired can not only obtain location information, but also know if there are obstacles in their way. The experiments show that the proposed system is reliable and can help the visually impaired in both indoor and outdoor environments. Furthermore, this study offers a new idea for guiding the visually impaired and provides important findings on system design, which might be useful for future research.

並列關鍵字

navigation ； vision-based ； visually impaired

參考文獻

[1] S. Agarwal, N. Snavely, I. Simon, S. Seitz, R. Szeliski: Building Rome in a Day. In: ICCV (2009)

[2] G. Brostow, J. Fauqueur, and R. Cipolla, “Semantic object classes in video: A high- definition ground truth database,” PRL, vol. 30(2), pp. 88–97, 2009.

[3] J. Bie, B. Visser, J. Matsari, M. Singh, T. van Hasselt, J. Koopman, B. Kröse, “Guiding the Visually Impaired through the Environment with Beacons”, UbiComp '16 Proceedings of the 2016 ACM International Joint Conference on Pervasive and Ubiquitous Computing: Adjunct Pages 385-388

[4] V. Badrinarayanan, A. Kendall, and R. Cipolla, “Segnet: a deep convolutional encoder- decoder architecture for image segmentation,” arXiv:1511.00561, 2015.

[5] D. Comaniciu, P. Meer, "Mean Shift: A Robust Approach Toward Feature Space Analysis", IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 24, no. 5, pp. 603- 619, May 2002.