基于数学形态学和神经元网络的货币识别

Currency Recognition Using Mathematical Morphology and Neural Networks

基于Ｇｒａｎｕｌｏｍ ｅｔｒｉｅｓ 定义的基础，给出了模式谱的定义．运用基于“开”运算的Ｇｒａｎｕｌｏｍ ｅｔｒｉｅｓ得到纹理的模式谱，并以此作为纹理的特征向量，通过双隐层的人工神经网络分类器进行分类，达到识别的目的．给出了该方法在货币识别上的应用实验结果．

As m orphological im age analysis tools granulom etrics are particularly useful for estim ating objectsizes in binary and grayscale im ages,or characterizing textures based on their pattern spectra(i.e., granulom etric curves).In this paper,the pattern spectrum of texture based on grayscale granulom etries w hich use opening operators w ith line structures was intruduced.The pattern spectrum w as used as the feature of texture and the tw o hidden layer neural networks w ere used for recognition.At the end, experim entalresults were shown.