Application Research Of Remote-sensed Image Classification Using Algorithm-improved BP Neural Network

The rapid development of remote sense technology provides us a lot of timely and efficient data,but how to extract useful information from these dataset is a important problem, among these problem, the clsssification of the image—being an important facet of information extract—can not reach the precision demand using the Bayes-based classificaion machine, but the neural network based on non-linear mapping provides a better solution to the classification problem.This article studied the application of BP neural network and its improved algorithm on the classification of remote-sensed image, the structure of the article is:First, the paper introduced the progress of neural network and classification of remote-sensed image.Second, the theory foundation origin, structure and character of neural network been discussed. After that, BP neural network and its L-M improved algorithm was introduced. Combined with the characteristic of neural network and the outcome of forerunners, the article studied the possibility and existing problem of using neural network to classify remote-sensed image.In the end, the L-M-based BP neural network was constructed, and the network was used to classify the remote-sensed image of Luoyugou in Tianshui region, then, the principal component transform was conducted and result in more rapid convergence velocity of the trianning time of the network and more accuracy ofclassification.