數字字符識別廣泛的應用到汽車牌照識別,大規模數據統計,財務、稅務等金融領域和郵件分揀等領域中。隨著國家信息化進程的加速,手寫數字識別的應用需求將越來越廣泛,因此應當加強這方面的研究工作。數字字符識別的方法有很多種,比如基于基于最小錯誤率Bayes決策和最小風險的Bayes決策,基于獨立分量分析,人工神經網絡等方法。
本論文設計是基于最小錯誤率Bayes決策和最小風險的Bayes決策的手寫數字字符分類。在Visual C++ 6.0的環境下,利用MFC開發出模擬手寫環境,通過對手寫數字字符的位置定位及其特征的提取,并利用基于最小錯誤率Bayes決策或最小風險的Bayes決策相關的理論知識,計算出相應判別函數和損失函數的值,并實現對0到9模擬手寫字符的分類。
關鍵字:Visual C++ 6.0,Bayes決策,數字字符識別。
Abstract
Figure recognition is widely used in license screening, large scale data analysis, financial and tax fields and mail sorting. With the acceleration of information development, handwriting figure recognition is in great need, and related research should be stressed. There are many methods of figure recognition, such as Bayes decision of minimal false rate and Bayes decision of least risk, based on individual part analysis, and artificial intellectual network.
The design included in this essay is based on figure recognition of bayes decision of minimal false rate and that of least risk. In the environment of Visual C++ 6.0, we use MFC to develop mimic handwriting situation to get the result of discrimination function and loss function and to realize the categorization of figures 0 to 9 through locating handwriting figures and the traction of related characteristics as well as bayes decision of minimal false rate and least risk.
Keywords:Visual C++ 6.0, Bayes decision, Figure recognition