基于遗传神经网络的零件图像非线性校正

提出了基于遗传神经网络校正非线性失真图像的方法。首先，用遗传算法优化神经网络的权值，构成遗传神经网络。然后，从标准的矩形栅格的失真图像中提取特征样本，样本的坐标用于训练遗传神经网络，标准矩形栅格中的样本的坐标作为目标输出。最后，以失真图像所有像素的坐标作为遗传神经网络的输入；其输出的坐标经过灰度级插值，实现图像的非线性校正。实验结果表明，文中提出的方法是有效的。

Nonlinearity Rectification of Part Image Based on Genetic Neural Network

A method for rectifying nonlinearity geometric distorted images based on genetic neural network is presented. Firstly, the genetic algorithm is used to optimize the weights of neural network and the genetic neural network is constructed. Then, the sample coordinates are extracted from a geometric distorted rectangular grid and are used to train the genetic neural network, and the sample coordinates from the original rectangular grid are used as the desired output values. Finally, all pixel coordinates from a distorted part image are regarded as the inputs of the genetic neural network and the gray level for all pixel coordinates of genetic neural network output is determined by gray-level interpolation to obtain non-linearity rectification of part image. Experimental results show that the method is effective for nonlinearity rectification of the part image.