Artificial Neural Network-based Remote Sensing Image Classification

With the space remote sensing technology continuously improved, remote sensing image spatial resolution is also rising. People can get more and more useful data and information from remote sensing images. So it is an important means to obtain the spatial information from remote sensing images by classifying the remote sensing images. However the spectrum values of remote sensing images are mixtures of various natural features, so there often exist that different natural features have the same spectrum value or different spectrum values are represented as the same natural feature. Because of these, it is very difficult to only rely on the pixel spectral similarity between each other to raise the whole classification accuracy. On the other hand, artificial neural network have nonlinear characteristics and more strong fault-tolerant capabilities, so to solve problems above can be possible by using artificial neural network.This paper reviews the research background of remote sensing image classification, a brief overview of the remote sensing image classification concepts and principles, discussed in detail of the traditional classification of remote sensing methods--supervised and unsupervised classification, and the emergence of new classification in recent years the artificial neural network method. The principle of various network models, algorithms and their respective advantages and disadvantages of a comparative analysis. By using C#.NET and ArcEngine technology, and call the MATLAB engine to achieve artificial neural network-based automated classification of remote sensing images. By analyzing the classification errors appear in the process sub-phenomenon, put forward in the classification process to add slope factor method, and to combine the non-remote sensing data and remote sensing data. The training process in the network model to sum up the relevance of slope and surface features, and the final threshold of network nodes and weight stored in the trained network. Through the latter part of the classification evaluation showed that using this method makes the classification accuracy have been significantly improved. Finally, the use of different network models and algorithms classification of the actual image classification, as well as analysis and comparison of classification results, and choose the best network model. The results show that, artificial neural network classification in classification performance is superior to the traditional classification method, the best performance BP network model is the training strategy based on an early end to the Levenberg-Marquardt method.In short, remote sensing image classification in pattern recognition is a more complicated issue. Remote sensing image classification of supervision and non-supervised classification methods is the most basic and general method in image classification. Traditional supervised and unsupervised classification despite their different strengths, but there is some deficiency. And the new classification methods such as neural networks with adaptive, self-learning, associative memory storage and distribution of good character, is that people attach importance to and is widely used in image classification. To break the traditional method of statistical classification of limitations, and improved the speed and accuracy of classification. Although a variety of classifications have their own characteristics, but in practice also needs to integrate and apply multiple classification methods, to improve the classification accuracy and precision.