首页>
外国专利>
Procedure to evaluate signal of electronic image sensor during pattern recognition of image content of test piece, by forming characteristics, making fuzzy, interference, making defuzzy and deciding class membership
Procedure to evaluate signal of electronic image sensor during pattern recognition of image content of test piece, by forming characteristics, making fuzzy, interference, making defuzzy and deciding class membership
Two dimensional image of local space of window of n x n pixels (2) is transformed into two dimensional image in frequency space. Frequency spectrum spectral amplitude characteristic values, are selected (09). Membership of each spectral amplitude value to a characteristic is determined using fuzzy membership function (13). Concrete membership (18) is determined by making defuzzy and is compared with predetermined threshold value (21) for classification (19). Grid of N x N windows (01) is laid over entire image to be analyzed, each window of n x n pixel (02). Two dimensional image of local space is transformed into two dimensional image in frequency space. Its frequency spectrum is formed by spectral coefficients. Sum of spectral values is formed (07), spectral amplitude values are the characteristic values. Characteristics that are typical for the image content are selected (09). Membership of each spectral amplitude value to characteristic is done by weighting using fuzzy membership function. Higher ranking function is created (16) by subjunctive conjunction of all membership functions (13) of characteristics (11). Concrete membership or sympathy value (18) is determined from functions (16) by making defuzzy. Sympathy value is compared with predetermined threshold value (21) for classification (19).
展开▼
机译：将n x n像素的窗口局部空间的二维图像（2）转换为频率空间中的二维图像。选择频谱频谱幅度特征值（09）。使用模糊隶属度函数（13）确定每个频谱幅度值对特性的隶属度。通过模糊处理确定具体的隶属度（18），并将其与预定的阈值（21）进行比较以进行分类（19）。 N x N个窗口（01）的网格放置在要分析的整个图像上，每个窗口n x n像素（02）。将局部空间的二维图像转换为频率空间的二维图像。它的频谱由频谱系数形成。形成频谱值的总和（07），频谱振幅值是特征值。选择图像内容的典型特征（09）。通过使用模糊隶属函数加权来完成每个频谱幅度值与特征的隶属关系。通过特性（11）的所有隶属函数（13）的虚拟连接，可以创建（16）更高级别的函数。具体的隶属度或同情值（18）通过使功能模糊而从功能（16）中确定。将同情值与预定阈值（21）进行比较以进行分类（19）。
展开▼
