- 學位論文
植基於色彩及紋理分析之鈔票偵測系統
A Bill-Detection System Based on Color and Texture Analyses
摘要
因數位影像技術不斷提升，製造偽鈔變得更容易且更便宜。為了解決偽鈔問題，本論文主要目的在發展一套鈔票偵測系統(Bill Detection System，即BDS)，此系統將於掃描影像或列印時偵測數位影像中是否包含鈔票，降低偽鈔製造者利用高解析度的數位影像產品來製作偽鈔的機會。本系統分成前置處理、色彩分析及紋理分析三個主要步驟。本論文使用132張含有美金鈔票及50張含日幣之影像做測試。研究結果顯示，本系統美金部分辨識率達84.8%，日幣部分辨識率達84%。總結而言，未來可將這套系統植入影像相關的電腦週邊設備上(如：掃描器、印表機或多功能事務機)，於一開始掃瞄鈔票或是在影印鈔票時防堵讓偵測系統功能更趨完善。
並列摘要
Because of continuous improvement of today’s digital imaging technology, the task to create counterfeit with high resolution has become easier and cheaper. To solve the counterfeit problem, the objective of this research was to develop a bill detection system (BDS) that could be used to detect digital images and to determine if the image includes a banknote. Our system was designed using pre-processing, followed by color and texture analyses. To test our system, a database of 132 digital images with U.S. banknotes and 50 digital images with Japanese banknotes were collected. As a result, our system has achieved a detection rate of 84.8% for the U.S. banknotes and 84% for the Japanese banknotes. In summary, our system could be potentially embedded in image related peripherals (e.g., scanners, printers, or multi-function peripherals) to prevent counterfeit from being printing or scanning.