Application and improvement of SOM network in remote sensing image classification
SOM神经网络改进及在遥感图像分类中的应用*

Contrary to the characteristics and weak point of self-organizing maps neural network, this paper shows a SOM which is improved by genetic algorithm, and improves the method to classifying remote sensing image based on self-organizing mapping network through the input vector, selecting the number of competitive layer neurons and the initializing weight vector. Finally, the method is used to classify an ETM+ satellite remote sensing images of Xi an. It is validated that through the improved self-organizing feature map network based on genetic algorithm, classification of remote sensing image have higher accuracy and efficiency than traditional self-organizing feature map network. The SOM based on genetic algorithm is easy to be achieved, and has practical value.