Data-Matrix 二維條碼辨識的性能改善方法

NCUIR > College of Electrical Engineering & Computer Science > Executive Master of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/74631

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/74631

Title:	Data-Matrix 二維條碼辨識的性能改善方法
Authors:	杜國綸;Tu, Kuo-Lun
Contributors:	資訊工程學系在職專班
Keywords:	條碼辨識
Date:	2017-06-23
Issue Date:	2017-10-27 14:34:26 (UTC+8)
Publisher:	國立中央大學
Abstract:	Data-Matrix 節省空間的特性，已經被廣泛的運用在汽車、航太、半導體與電子業零件，然而，目前市面上的 Data-Matrix 條碼讀取器，在進行條碼的解析時，若是遇到模糊、光照不均、扭曲等難以識別的條碼時，會使用多種影像前處理方法重複不斷的嘗試識別，若識別結果失敗，往往已經耗費不必要的計算時間。本論文基於目前 Data-Matrix 辨識主流的開源函式庫 LibDmtx，設計一個 Data-Matrix 二維條碼辨識系統。LibDmtx 演算法在條碼定位使用 Sobel 遮罩沿著邊緣的正切線與負切線求出條碼區域，在影像品質不佳或具有大量不必要的高頻資訊時會造成條碼識別正確率低落且耗費大量的時間，並且無法事先預測出該條碼影像根本無法識別造成時間的浪費。本研究針對此一缺陷，加入一個創新的分類器，用以事先預測該條碼影像是否可以成功的識別，藉此避免部分低品質條碼影像從事冗長的影像前處理與後續條碼識別步驟，節省系統整體運行時間。此外，在影像前處理階段，我們應用離散小波轉換來增強影像品質，可以改善條碼辨識率，達成系統速度與正確率兼顧的目標。;Data-Matrix two-dimensional barcodes present extensive information within a compact area. Data-Matrix barcodes have been widely used in automobile, aerospace, semiconductor, and electronic components. However, the Data Matrix barcode readers currently available on the market use multiple image preprocessing methods for identification of barcodes that appear fuzzy, unevenly lit, distorted, and otherwise unrecognizable; with such systems, failed identification often needlessly wastes computational time. Poor image quality or a large amount of unnecessary high-frequency information can hinder the barcode recognition and consume a substantial amount of time. In addition, unrecognizable barcodes cannot be predicted in advance, resulting in wasted time. This study presents a Data-Matrix two-dimensional barcode recognition system based on the mainstream open source software libdmtx, which uses Sobel filter edge detection to obtain the barcode along the positive and negative tangent lines. To address the shortcomings of typical systems, a novel classifier is applied in this study to predict whether a given barcode can be successfully recognized. This prevents cumbersome image preprocessing and subsequent recognition steps for low-quality barcode images, thereby saving computational time. Furthermore, during the image preprocessing phase, a discrete wavelet transform approach is employed to enhance image quality and improve barcode recognition, thereby improving both speed and accuracy of system operations.
Appears in Collections:	[Executive Master of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	614	View/Open

社群 sharing

Loading...