Back to results list
Please use this identifier to cite or link to this item:
|Title:||Improvement and estimation of classification accuracy for remotely sensed images||Authors:||Mao, Haixia||Keywords:||Hong Kong Polytechnic University -- Dissertations
|Issue Date:||2010||Publisher:||The Hong Kong Polytechnic University||Abstract:||A land use inventory provides the means from which data can be extracted to enable such as land management and the decision making necessary to promote sustainable environmental development, where image classification is one of the most important procedures. The overall aim of the study in this thesis is to improve the reliability of the land use inventory, by specific focus on the development of methods, to improve and estimate the classification accuracy of remotely sensed images. The following research objectives are thus identified: - to unmix the mixed pixels for hyper/multi-spectral remotely sensed images; - to propose a new multiple classifier system; - to propose a spatial sample strategy. Firstly, a solution of unmixing mixed pixel, including endmember extraction, abundance generation and sub-pixel mapping, for hyper-spectral images is proposed. One of the main contributions of this research is the proof for AMEE assumption and the proposal of the Improved AMEE to enhance the performance of endmember extraction. Pre-judgment before applying the Least Squares is carried out to calculate the abundances. To further improve the performance of pixel unmixing, Reliability-based Sub-pixel Mapping is proposed to locate each endmembers in a mixed pixel. Secondly, pixel unmixing for multi-spectral remotely sensed image based on single band is addressed. The Mountain Clustering is introduced to extract endmembers. The Grey Correlation method is still used to generate the abundance of each endmember. And the Improved Cellular Automata is proposed for sub-pixel mapping. As pixel unmixing is implemented in each single band, the Multiband Synthesis is proposed to integrate the result from partial overall. Thirdly, to improve the accuracy of land use classification, an Eigen-values based Multiple Classifier System is proposed in this study. A posterior probabilities matrix of each component classifier is obtained and the eigen-values are calculated based on the matrix. In Eigen-values based Multiple Classifier System, the eigen-values are used to weight the classifiers. With the proper correspondence between eigen-values and classifiers, the proposed method has proved to be effective for image classification land use inventory.
Fourthly, a sampling strategy based on error-distribution is proposed to assess the accuracy of the image classification results. The error surface is generated, based on image classification errors, where those errors are regarded as noise in the research field of signal processing. Firstly, after a approximate classification, a prior partition of the original spatial data, based on the different error levels, is found. Secondly, based on the assumption that the errors follow the normal distribution, different Gaussian filters are used to remove them in each region. Error-free spatial data can then be obtained. Finally, the difference between the original spatial data and the error-free data is used to build an error surface. Hence, the extreme points on the error surface are considered as the representative points of the sample and the least sample size is determined. In summary, mixed pixels are one of the main obstacles for image classification; the classification method itself has direct impact on the classification accuracy; the accuracy assessment system will influence on the evaluation of the classification result. If all the above problems can be solved effectively, a great success can be obtained in image classification. Consequently, this research focuses on above objectives and contributes to the quality of the land use inventory in the following two areas: (a) it improves image classification accuracy by solving the mixed pixel problem and thereby provides an improved multiple classifier system; (b) it contributes to the improvement of the reliability of the accuracy assessment result, by the means of a new spatial sample strategy.
|Description:||xxi, 174 p. : ill. ; 31 cm.
PolyU Library Call No.: [THS] LG51 .H577P LSGI 2010 Mao
|URI:||http://hdl.handle.net/10397/2864||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|b23745137_link.htm||For PolyU Users||162 B||HTML||View/Open|
|b23745137_ir.pdf||For All Users (Non-printable)||3.64 MB||Adobe PDF||View/Open|
Citations as of Jun 18, 2018
Citations as of Jun 18, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.