鈔票的大量流通,一定會有破損或是沾污的情形,為了減少人力資源的浪費及誤判,本研究建立一個基於DSP架構的高速點驗鈔系統。針對新台幣,做到面額、入鈔方向、折角以及新舊檢驗等功能。此點驗鈔系統由六大模組組成,包括餵鈔模組、厚度鑑別模組、磁性判別模組、影像辨別模組、出鈔模組以及使用者輸入界面模組。各個模組功能如下,餵鈔模組負責將鈔票帶入辨鈔機器中,厚度鑑別模組判斷鈔票的厚度是否正常,磁性判別模組則是判別鈔票的磁性,而影像辨識模組則是本研究的重點,負責一切有關於影像辨識的功能,送鈔模組是依判別結果將鈔票送至不同的出鈔槽,使用者介面則是處理參數的設定、輸出辨識結果。影像辨識模組中,鈔票影像透過Contact Image sensor(CIS)讀入由ADI所生產的ADSP-BF533 DSP系統中,利用鈔票特徵來進行辨識。實驗結果針對無折損的鈔票,面額及入鈔分向部分可以達100%,折角部分檢測也可以達到 96%,系統設定為每分鐘可辨識720張鈔票。
The circulating currencies, there will be damaged or soiled situations. In order to reduce the waste of human resources and misjudgment, this study established a banknote recognition system based on DSP architecture to do denomination, into banknote direction, angular detection, and the old and the new banknotes recognition for NT dollars. There are six modules in this banknote recognition system, including feeding module, thickness of discrimination module, magnetic of discrimination module, image recognition module, delivery module and user interface module. The feeding module is in charge of sending banknotes to the system. Thickness discrimination module is in charge of detecting the thickness of the banknotes. Magnetic discrimination module is in charge of detecting the magnetism of the banknotes. Image recognition module is the key point of this study, and is in charge of image recognition. Delivery module will send the banknotes to different slots based on detection result. And user interface module is in charge of parameters setting and display the detection result. The experimental results in banknote denomination and direction with the sample banknotes used in this experiment can reach 100%, the angular detection can reach 96%. This system is configured to recognize the 720 bills per minute.