基于径向基函数神经网络的栅格图像非线性畸变校正

为了减小光电成像测量系统中存在的非线性畸变，提高测量精度，提出了一种基于径向基函数神经网络的图像畸变校正方法。提取带有桶形畸变的栅格图像中的栅格交叉点作为控制点，利用光学成像关系推算出栅格交叉点的理想无畸变位置，构成径向基函数神经网络的训练集。经过训练，可以确定径向基函数神经网络结构的优化参数。针对栅格图像进行了畸变校正实验，并与多项式变形法进行了比较。实验结果表明，所提方法能够自动、有效地校正图像畸变，效果优于多项式变形法。

Nonlinear Distortion Correction of Grid Image Based on Radial Basis Function Neural Network

In order to decrease nonlinear distortion of electro-optical imaging measurement system, a distortion correction method based on radial basis function neural network is proposed to improve measure precision. Cross points of the black lines can be found and regarded as control dots by edge detection and thinning of a grid image with barrel distortion. According to imaging characteristic of an optical system, coordinates of cross points in an undistorted image can be calculated from ones in the distorted grid image. With control dot pairs, a training set of radial basis function neural network can be set up. Optimal structural parameters of the radial basis function neural network can be obtained by training. The proposed method is tested and compared with a polynomial warping method. Experimental results show that the proposed method can correct distortion automatically and efficiently, and has a better distortion correction than the polynomial warping method.