同步定位與地圖建構結合磁場地圖提升移動式行動裝置室內定位精度之研究

The Study of Simultaneous Localization and Mapping Integrating Magnetic Field Maps to Improve Indoor Positioning Accuracy of Mobile Devices

李尚桀(Li Shang-Jie)

指導教授 ： 甯方璽

國立政治大學/社會科學學院/地政學系/碩士(2022年)

https://doi.org/10.6814/NCCU202201193

摘要

隨著科技不斷的進步，人們對移動式行動裝置的使用量大幅增加，利用其對室內外進行定位與導航的需求也逐漸成長，如何達成更高的定位精度，已是各研究致力於發展的目標。對室外定位而言，全球導航衛星系統的發展已讓室外定位性能趨近完善，然而受限於訊號遮蔽影響，無法有效的對室內使用者進行定位，因此室內定位技術的發展成為近年研究的方向。目前常見的室內定位技術包括，無線射頻、影像視覺及行人航位推算，但各技術的優缺點，使得目前沒有單一室內定位技術能夠解決各項環境因素。
鑒於移動式行動裝置的普及以及運算能力的發展，已能夠大量的處理資料以及快速的分析數據，本研究將以智慧型手機作為實驗裝置，利用移動式行動裝置本身內建之相機與磁力感測器，對室內環境中的特徵與建物結構中磁場之影響量進行資料收集。透過事先建立之室內磁場指紋地圖，以WKNN匹配法獲取初始位置；結合視覺同步定位與地圖建構，利用ORB特徵以推算使用者室內坐標，最後利用耦合及磁場約制將視覺同步定及磁場定位成果結合，並進行精度分析，由研究結果顯示，單一定位技術之定位精度僅1.5至2 m，經耦合並利用磁場確定起始點及約制，精度可達0.5至0.7 m，不同廠牌型號行動裝置以本研究之所提出之方法亦可達到相同之精度。

關鍵字

室內定位 ； 移動式行動裝置 ； 同步定位與地圖建構 ； 磁場定位

並列摘要

With the continuous advancement of technology, the use of mobile devices has increased, and the demand for using them for indoor or outdoor positioning has also gradually grown. To achieve higher positioning accuracy has been the goal of various researches. For outdoor positioning, the development of GNSS has improved the outdoor positioning performance. However, due to the influence of signal shielding, it cannot effectively locate indoor users. Therefore, the development of indoor positioning technology has become a target of research in recent years. At present, common indoor positioning technologies include, radio frequency, image vision, and pedestrian dead reckoning. Each technology has its pros and cons which cannot perfectly resolve the issues of indoor positioning technology since various environmental factors would have an impact on it.
In view of the development of mobile devices and computing power, it has been able to process a large amount of data and analyze it rapidly. Therefore, our research will use a smartphone as an experimental device. Using the built-in camera to collect data on the characteristics of the room surroundings, and the magnetometer to detect the influence of the building structure on the magnetic field. First, obtain the initial position by the WKNN matching method through the indoor magnetic field fingerprint map established in advance, and combine the visual simultaneous localization and mapping, the ORB feature is used to calculate the user's indoor coordinates. Finally use the coupling methods to combine the two positioning results, and perform a precision analysis. The research results show that the positioning accuracy of a single positioning technology is only 1.5 to 2 m. After coupling and using the magnetic field to determine the starting point and constraint, the accuracy can reach 0.5 to 0.7 m. Different brands of mobile devices can also achieve the same accuracy by the method proposed in this study.

並列關鍵字

Indoor positioning ； Mobile devices ； Simultaneous localization and mapping ； Magnetic field positioning

參考文獻

一、 外文參考文獻

Andrew, A. M. (2001). Multiple view geometry in computer vision. Kybernetes.

Beer, J., Blakemore, C., Previc, F. H., & Liotti, M. (2002). Areas of the human brain activated by ambient visual motion, indicating three kinds of self-movement. Experimental brain research, 143(1), 78-88.

Bay, H., Tuytelaars, T., & Van Gool, L. (2006, May). Surf: Speeded up robust features. In European conference on computer vision (pp. 404-417). Springer, Berlin, Heidelberg.

Birsan, M. (2010). Recursive Bayesian method for magnetic dipole tracking with a tensor gradiometer. IEEE Transactions on Magnetics, 47(2), 409-415.