| As an important branch of remote sensing technologies, the image classification has been widely regarded by researchers. In this paper, combined with the characteristics of the remote sensing image and the knowledge of the fussy technology and the neural networks, I made a classification experiment to remote sensing image using a kind of the improved fuzzy neural networks, compared with the traditional methods of the remote sensing image classification. The remote sensing image is mostly made of the mixed pixels, but the traditional methods of the remote sensing image classification are based on pure pixels that aren't according to the reality. The application of fuzzy technology can be good to resolve the mixed pixels of the remote sensing image .The Neural network has the characteristics of self-learning, self-organizing, parallelism, etc which is good idea to the great data of the remote sensing image. The remote sensing image classification has a very big development potential in the combine of the fuzzy technology and the neural networks. There are two kinds of the combine of fuzzy technology and neural network currently .one kind is combined of the fuzzy logic and the feedback neural network .The other is combined of the fuzzy logic and feed forward neural network. Because the former has already been researched widely, so I carry on a classification research to the latter, that is combined of the fuzzy logic and the Back-Propagation Network (BP) .At the same time, I made classification experiments with the methods of the Back-Propagation Network (BP) and the fuzzy c-means clustering (FCM) compared with this kind of method .The result show that this kind of method is more vivid than the other above methods. But in the meantime, there are a lot of problems need to be resolved in the classification process. Over all, the combine of the fuzzy technology and the neural networks is effectual methods in remote sensing image application that deserve being researched on more. |
PDF Full Text Request