Loading...

Texture Change Detection in Hyperspectral Images

Dianat, Rouhollah | 2010

520 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 41163 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Kasaei, Shohreh
  7. Abstract:
  8. In this thesis, change detection in hyperspectral images is investigated. In this process, two hyperspectral images captured from the same scene but in different time instances are given and we intend to detect the occurred changes in the scene. A specific change detection algorithm contains four different steps; namely, preprocessing, selection of the criterion, postprocessing, and decision making. Dimension reduction as a critical process is also deeply investigated to be performed before classification (for decision making purposes.) Two new methods are proposed in the thesis. The proposed MCRD method is designed for dimension reduction and the MPR method is related to the change detection principle. MCRD is constructed so that it holds all useful properties of the PCA method (as the most well-known dimension reduction approach) while it overcomes the shortcoming of PCA by incorporating the neighborhood relations among image points. In MCRD, a general framework based on quadratic equation construction is introduced to account for spatiality. This generality is achieved by embedding a matrix called feature extraction. Two versions of this matrix are developed in this thesis. The first one takes texture contrast into account (a spatial feature), and the second one uses the class separability concept to construct the matrix. The MPR is an improvement to the well-known CPR change detection method. Same as MCRD, in MPR, by using the feature extraction matrix, we have managed to embed the spatial information into CPR to obtain better results. In the experimental result section, various quantitative and qualitative experiments are given to show the superiority of both proposed methods. The results show that MCRD outperforms PCA about 10% in classification accuracy; MPR has better performance up to 20% in overall error; and the system combined with MCRD and MPR improves the performance over the original PCA and CPR-based methods about 13%
  9. Keywords:
  10. Dimension Reduction ; Remote Sensing ; Change Detection ; Hyperspectral Images

 Digital Object List

 Bookmark

No TOC