论文部分内容阅读
本文提出了数学形态学算子的前馈神经网实现的模型.各种形态学算子(形态交离、腐蚀、膨胀、开启、闭合及多结构元素的基本形态运算的交或并)都可以根据算子本身置定权值、阈值和网结构以前馈神经网来实现.相应于对应算子和结构元素的神经网可通过简单的训练而直接得到.同时,我们也讨论了上述各种算子的布尔函数实现.
In this paper, a mathematical model of feedforward neural network is put forward, in which all kinds of morphological operators (the intersection or union of morphological operators, corrosion, expansion, opening, closing and multi-structure elements) According to the operator itself set the weights, thresholds and network structure to feed neural network to achieve. Corresponding to the corresponding operators and structural elements of the neural network can be obtained through a simple training and direct, and we also discussed the above calculations Boolean implementation of children.