本研究以統一發票對獎當做欲辨識的對象，以模式辨認(Pattern Recognition)為基礎發展出快速且準確率高的辨識系統。不採用類神經網路是因為其需要大量的樣本以作事前的學習才能達到有效的辨識率，而本文所使用之方法比之所需要的樣本數較少且辨識效率也可達所需求，再者發票上的號碼是阿拉伯數字且為鉛字印刷類型，並不需要使用到太過複雜的方法。在發展過程中發現模式辨認的辨識效率雖好，卻無法克服圖像旋轉的問題，因此最後利用發票旋轉後之發票號碼的傾斜角求出旋轉之角度，再將圖片轉正後再透過模式辨認來辨識號碼，經過傾斜角度由50°到-49°等100 張不同角度之圖片的測試與使用實際發票200 張測試證實此方法確實可行。

In order to develop a fast and high accuracy of the recognition system, we used the pattern recognition to identify the uniform invoice in this paper. The artificial neural network is not used in the paper, because a large number of samples must be learned in advance to achieve the effective recognition rate. The approaches, we propose, only need a few of samples and can meet the effective recognition rate. Furthermore, the invoice numbers of the font structure belong to printing type, there is
no need to use the complicated method.In the recognition process, we found that the pattern recognition has a good recognition effect, but
can not overcome the problem of rotated image. Therefore, we proposed a new method to deal with above problem. First, the angle of the rotated invoice number image is calculated and turned back the rotated invoice number image by image processing techniques. Then, the template matching is performed to identify invoice numbers. Finally, we tested 100 rotated images from 50° to -49° and 200 real invoices. Experimental results show that the proposed method of the paper is feasible.