請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/72281| 標題: | 使用不可視光偏振圖形之室內定位系統 Indoor Positioning System using Invisible Polarized Light Pattern |
| 作者: | Wei-Nin Chang 張惟甯 |
| 指導教授: | 蔡欣穆(Hsin-Mu Tsai) |
| 關鍵字: | 偏振光,慣性感測器,粒子濾波器,室內定位, Polarized Light,Inertial Sensor,Particle Filter,Indoor Positioning, |
| 出版年 : | 2018 |
| 學位: | 碩士 |
| 摘要: | 慣性測量單元 (Inertial Measurement Unit, IMU)長久以來因為感測器所收到的雜訊會累積的非常快速而有偏移的問題,進而造成在定位追蹤時會有很大的誤差。近幾年來有許多不同方法來校正IMU,但有些方法需要花很多時間做指紋識別 (fingerprinting),有些需要很多額外的硬體,或是有些定位的準確度很低。我們的論文提出一個不需要對一般的燈具做任何調整,即可以投射出光偏振的圖形。這個圖形在不同格子上,有著不同的“偏振顏色”,這些顏色是人眼所看不到的,但一般商用手機的相機可以偵測到這些顏色。此方法可以提供很精細的標記去校準IMU的偏移。我們使用貼有雙折射特性的透明膠帶的偏振片作為一般燈具的燈罩,並利用不同結構的膠帶對IMU提供簡單的限制,接著運用粒子濾波器 (particle filter)結合IMU與顏色來定位。我們的論文提供了低耗能、低成本、不需要過多額外的硬體且高準確度的定位系統。我們的實驗結果顯示出定位誤差的中位數為0.2公尺,而且轉向的改變也不會影響我們定位的準確度。 Large drift over time has been a significant problem for localization using Inertial Measurement Unit (IMU) for a long time, as noises within the sensor output accumulate quickly and result in large tracking errors. Recent works use a number of methods to calibrate in order to mitigate the problem. However, they either require fingerprinting, redundant infrastructures, or has low accuracy. Our thesis proposes an novel system to augment existing luminaries, such that it projects polarized light pattern. This pattern exhibits different ’polarized colors‘ at different spatial locations, which are not perceptible by human eyes, but can be picked up by the commodity camera of a mobile device. The pattern can serve as fine-grained landmark to constrain the tracking result and correct IMU’s drift. Exploiting the birefringence optical property of transparent tapes placed on polarizer, we create a light cover to augment existing luminaries. With different configurations of tape layers within the light cover, the projected pattern can be used to restrict the tracking result and provide better accuracy. Particle filter is used to fuse the color data and the IMU sensor data and produce the location estimation. Our work provides a low-power, low-cost, and high accuracy solution for indoor positioning and do not need additional infrastructure. Experiments show that the system produce localization median error as low as 0.2 meters and the tracking is robust against orientation changes. |
| URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/72281 |
| DOI: | 10.6342/NTU201803755 |
| 全文授權: | 有償授權 |
| 顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
| 檔案 | 大小 | 格式 | |
|---|---|---|---|
| ntu-107-1.pdf 未授權公開取用 | 7.79 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。
