Identifier

etd-11092007-160127

Degree

Doctor of Philosophy (PhD)

Department

Geography and Anthropology

Document Type

Dissertation

Abstract

Recent research techniques, such as genetic algorithm (GA), Petri net (PN), and cellular automata (CA) have been applied in a number of studies. However, their capability and performance in land-cover land-use (LCLU) classification, change detection, and predictive modeling have not been well understood. This study seeks to address the following questions: 1) How do genetic parameters impact the accuracy of GA-based LCLU classification; 2) How do image parameters impact the accuracy of GA-based LCLU classification; 3) Is GA-based LCLU classification more accurate than the maximum likelihood classifier (MLC), iterative self-organizing data analysis technique (ISODATA), and the hybrid approach; 4) How do genetic parameters impact Petri Net-based LCLU change detection; and 5) How do cellular automata components impact the accuracy of LCLU predictive modeling. The study area, namely the Tickfaw River watershed (711mi²), is located in southeast Louisiana and southwest Mississippi. The major datasets include time-series Landsat TM / ETM images and Digital Orthophoto Quarter Quadrangles (DOQQ’s). LCLU classification was conducted by using the GA, MLC, ISODATA, and Hybrid approach. The LCLU change was modeled by using genetic PN-based process mining technique. The process models were interpreted and input to a CA for predicting future LCLU. The major findings include: 1) GA-based LCLU classification is more accurate than the traditional approaches; 2) When genetic parameters, image parameters, or CA components are configured improperly, the accuracy of LCLU classification, the coverage of LCLU change process model, and/or the accuracy of LCLU predictive modeling will be low; 3) For GA-based LCLU classification, the recommended configuration of genetic / image parameters is generation 2000-5000, population 1000, crossover rate 69%-99%, mutation rate 0.1%-0.5%, generation gap 25%-50%, data layers 16-20, training / testing data size 10000-20000 / 5000-10000, and spatial resolution 30m-60m; 4) For genetic Petri nets-based LCLU change detection, the recommended configuration of genetic parameters is generation 500, population 300, crossover rate 59%, mutation rate 5%, and elitism rate 4%; and 5) For CA-based LCLU predictive modeling, the recommended configuration of CA components is space 6025 * 12993, state 2, von Neumann neighborhood 3 * 3, time step 2-3 years, and optimized transition rules.

Date

2007

Document Availability at the Time of Submission

Secure the entire work for patent and/or proprietary purposes for a period of one year. Student has submitted appropriate documentation which states: During this period the copyright owner also agrees not to exercise her/his ownership rights, including public use in works, without prior authorization from LSU. At the end of the one year period, either we or LSU may request an automatic extension for one additional year. At the end of the one year secure period (or its extension, if such is requested), the work will be released for access worldwide.

Committee Chair

Nina Lam

DOI

10.31390/gradschool_dissertations.2433

Share

COinS