用于图像识别的新型神经元网络

提出了一种用于图像识别的映封正交神经元网络。

在一般情况下待识别的样本空间的维数相当大,

为了有效地进行识别,必须把样本空间的维数降下来。目前常用的方法就是特征提取法,本文采用映射正交过程把样本空间映射成正交分类空间,并在此基础上,采用网络进行图像分类。计算机上模拟结果表明此网络县有时缺损和噪声图像进行正确识别的能力-Image recognition for a new type of neural network used in a video image recognition of the orthogonal neural network closure. In general, the sample space to be recognition of the considerable dimensions, in order to effectively identify, must be the dimension of sample space down. Commonly used method at present is the feature extraction method, this paper, the process of mapping the sample space orthogonal mapping classification into orthogonal space, and on this basis, the use of network image classification. Computer simulation results show that the county and sometimes the network image defects and noise of the ability to correctly identify