行動QR碼影像辨識演算法之設計

Abstract

近年來半導體產業及科技的進步，自動資料收集之技術已成為一值得投入之領域。隨著內建相機功能之行動電話的普及，越來越多科技應用相繼而生。針對此趨勢，本文提出一對於二維行動條碼-QR Code的相關研究，進而設計其解碼平台所需之影像辨識演算法，以便用於行動裝置中。藉由內建相機功能之行動裝置與無線網路的蓬勃發展，希望未來能利用此技術建構一可靠且快速之產品履歷系統。 本文提出之方法為利用手持式行動裝置配合內建的相機功能做為資料來源的擷取裝置，由外部讀取一個二維行動條碼影像加以辨識其中內含的資訊，藉由此流程達成快速與高容量之資料傳達與索引的功能。另外，目前相關研究尚無人提及扭曲條碼影像的辨識方法與相關的校正處理，於是本文提出一演算法能同時處理扭曲與非扭曲的QR Code影像，並加以辨識。本文主要的研究內容包含二維行動條碼研究、影像擷取與前處理、影像辨識演算法、扭曲圖形辨識與校正、行動裝置平台應用。最後，實驗結果證明本文提出之演算方法具有高度的強健性與準確率，能有效地辨識QR Code影像內含的資訊。

With the advancement of semiconductor technology and industry, the technology of automatic data collection has become a noticeable application field and worthy investment in recent years. In the meanwhile, more applications have been created on the mobile phone with embedded camera. To meet the trend, this thesis investigates QR Code technology, which is one of the main streams of mobile 2D barcode. An efficient image recognition algorithm is proposed to extract QR Code information using the camera of mobile phone. Along with the fast booming of mobile phone with camera function and ubiquitous wireless network, the proposed approach can provide a reliable and fast Electronic Pedigree System (EPS) in the future. This study utilizes the camera in the mobile device to capture a QR Code image and a program is developed to recognize and decode the image. The proposed QR Code reading procedure has faster information transmission ability and high data indexing capacity. In practice, QR Code images are often crooked and there are few researches that investigate the crooked image recognition. To solve this problem, this thesis proposes an algorithm that can recognize both crooked and flat QR Code image. The contents of this research include image capture and pre-processing, image recognition, and crooked image recognition and rectification of 2D barcode. Performance evaluation for the proposed algorithm has been conducted. The experimental results show that the proposed algorithm is robust and accurate, and can extract the QR Code information effectively.