基于GA-BP算法的多分辨率遥感影像融合技术

由于Landsat-5唯一的热红外波段遥感影像TM6的空间分辨率不高,使得其应用与研究程度远不及其它波段广泛。为此,运用GA-BP算法来提高TM6遥感影像的空间分辨率,并进行仿真实验,结果表明:①GA-BP算法有效地避免了BP算法陷入局部最小点、收敛速度慢的问题,是一种快速、可靠的方法。它的快速算法对数据量巨大的遥感图像更具实用价值。②从提高TM6遥感影像空间分辨率的仿真结果来看,无论计算效率还是遥感影像的融合效果,GA-BP算法都优于BP算法。③GA-BP算法既保留了TM6遥感影像的基本灰度分布信息,同时也提高了其空间分辨率,可以有效地运用到提高遥感影像空间分辨率的过程中。

A Study on A New Method of Multi-spatial-resolution Remote Sensing Image Fusion Based on GA-BP

Low spatial-resolution is a disadvantage of TM thematic remote sensing (TM6) image, which hampers its widespread application. So this paper put forward a new method based on the genetic algorithms and back-propagation algorithms (GA-BP) to improve the spatial resolution of TM6 image.
This method combines the genetic algorithms and neural network, replacing the BP algorithm with the genetic algorithm. Tests showed that:①GA-BP algorithm can avoid the problems of BP algorithm which indicate calculation procedure is easily caught in local minimal point and the convergence is slow. The advantages of GA-BP are especially of more practical use for mass calculation of remote sensing images.②According to the simulation result, GA-BP algorithm had a distinct advantage over BP algorithm, no matter calculation efficiency or fusion effectiveness.③GA-BP algorithm improves spatial resolution of TM6 image and the gray value can also be reserved, thus can be applied into the practice of improving the