基于RBF人工神经网络的遥感图像校正算法

遥感图像的几何校正是阻碍其应用的瓶颈问题。遥感图像的几何校正函数是一个非线性、不确定的复杂函数, 难以用精确的数学模型来表达问题, 对此提出了一种基于RBF 人工神经网络和地面控制点( GCP) 相结合的遥感图像几何校正算法。该算法利用RBF 网络能以任意精度逼近任意函数的特性, 模拟地表空间分布这一复杂的非线性函数。该算法原理简单, 易于实现, 是一种实用、可行的遥感图像几何校正算法。

Image Correction Using Interpolation Function

The geometric correction is a bottleneck problem which baffles the remote sensing images' application.Geometric rectification is a nonlinear,uncertain,complex dynamic function and hard to be described completely,so a method of remote sensing image geometric correction based on the RBF neural networks and Ground Control Point(GCP) is made in this paper.Firstly,the algorithm determines the control point used to correct.Then,use the characteristic of RBF neural network can represent all functions at any accuracy to simulated surface of this complex spatial distribution function of the nonlinear.From the experimental results,it comes to some conclusions: The algorithm's principle is simple,and easy to be realized.Usually,the algorithm has an error less than one-pixel,which is accurate enough for geometric correction.It is a kind of more practical algorithm of the remote sensing image correction.