Dissertation > Excellent graduate degree dissertation topics show

Based on Artificial Neural Network Classification of Remote Sensing Images

Author: LiYuanTai
Tutor: ZhaoJunSan
School: Kunming University of Science and Technology
Course: Cartography and Geographic Information Systems
Keywords: Remote sensing Image Classification Supervised classification Artificial Neural Networks MATLAB
CLC: P237
Type: Master's thesis
Year: 2010
Downloads: 237
Quote: 0
Read: Download Dissertation

Abstract

With the continuous development of space remote sensing technology, remote sensing image spatial resolution is also rising. People can get more from remote sensing images useful data and information, and for remote sensing image classification is an important means of access to information. Since the spectral values of remote sensing images feature a variety of mixed spectrum, there is a \The artificial neural network nonlinear characteristics and strong due to its fault-tolerant capabilities for solving the above problems may be provided. This paper reviews the research background of remote sensing image classification, a brief overview of remote sensing image classification concepts and principles discussed in detail the traditional remote sensing classification methods - supervised classification and unsupervised classification, and the recent emergence of some new classification - - artificial neural network classification. The principle of various network models, algorithms and their advantages and disadvantages are compared and analyzed. Through the use of C #. NET and ArcEngine technology, and calling MATLAB engine based on artificial neural network to achieve the automation of remote sensing image classification. By analyzing the classification process misclassification phenomenon occurs, the classification process proposed increase in the slope factor method, the non-remote sensing data and remote sensing data composite. In the network model training process of summing up the slope and correlation feature, the network node and the final thresholds and weights stored in the trained network. By the late classification evaluation shows that the use of the method makes the classification accuracy has been significantly improved. Finally, the use of different network models and algorithms for the actual remote sensing image classification, and classification results are analyzed and compared, selecting the optimal network model. The results show that the artificial neural network classification in the classification results superior to the traditional classification methods, in a variety of BP network model is based on the optimal performance early stop training strategy Levenberg-Marquardt method. In short, the remote sensing image classification is a complex field of pattern recognition problems, remote sensing image supervised classification and unsupervised classification method, the image classification is the most basic and summarized in two ways. Traditional supervised and unsupervised classification method, although their own advantages, but there are also some disadvantages. The new classification method such as neural networks for its adaptive, self-learning, associative memory storage and distribution of good characteristics, known to be valued and widely used in image classification, breaking the traditional limitations of the statistical classification method to improve the classification speed and accuracy. Although various classification methods have their own characteristics, but in practice also need comprehensive application of various classification methods to improve the classification accuracy and precision.

Related Dissertations

CLC: > Astronomy,Earth Sciences > Surveying and Mapping > Photogrammetry and Surveying, Mapping and Remote Sensing > Surveying, Mapping and Remote Sensing technology
© 2012 www.DissertationTopic.Net Mobile