被動式可見光標記之偵測與辨識

Abstract

隨著 LED 及通訊技術的發展，過去十年間可見光通訊（VLC）已 經引起越來越多關注。利用建築物中廣泛部署的照明系統作為傳輸媒 介，可見光通訊將可用於實作高準確度的室內定位系統。在近年的研 究中，可看到許多室內可見光定位（VLP）的方法，例如物體自定位 系統和基於標籤的定位系統。在物體自定位系統中，目標物須具備感 光元件，因此電源供應便是不可避免的問題，並且只有物體本身知道 自己的位置；基於標籤的定位系統則是利用相機拍攝環境中的定位標 籤，進而分析標籤相對於環境及相機的位置，以及解碼標籤中的身份 訊息，因此必須依賴較高複雜度的影像處理技術才得以辨識及解碼。 在本論文中，我們提出了一種使用被動式可見光標籤的定位系統，被 定位物體不需供電，並且使用光電二極體偵測反光標籤的反射光作為 定為媒介，不僅減少硬體設備的使用以及降低運算資源的需求，同時 也避免在定位過程中重建周圍環境的影像，因此適用於高度隱私要求 的應用情境。我們使用壓縮感知（Compressive sensing）快速偵測標籤 位置，搭配光柵掃描達到精準定位，最後使用細部標籤掃描解碼身份 訊息。實驗的結果顯示，定位精準度能達到 0.7 公分以內，而標籤樣 式的辨識準確度在 3 公尺的距離之下能達到 95% 以上的準確率。

With the advanced development of LED and communication technology, visible light communication (VLC) has gained more and more attention in recent decades. Relying on lighting system for transmission, it takes advantage of the integral facilities in buildings, and thus becomes a solution for accurate indoor positioning. Several indoor visible light positioning (VLP) methods have been proposed in recent studies, which can be categorized into self-positioning system and marker-based positioning system. Self-positioning system needs light sensing devices for target objects to receive light signal, therefore power is an inevitable problem. A marker-based positioning system relies on camera to capture image of fiducial marker attached to specific location or objects. High computational requirement for image processing cannot be avoided. In this thesis, we propose a VLP system that relies on the novel design of passive visible light marker to keep objects battery-less, and a single pixel light sensor to detect reflected light of markers, not only minimizes the computational power and facility, we can also avoid reconstructing the image of the environment for privacy-concerned applications. We leverage compressive sensing to support fast detection, followed by edge detection for high accuracy localization. And finally the marker scanning and identity decoding process. Our experiments show that the localization error is within 0.7 cm, and identity decoding accuracy can achieve higher than 95%.