WHDL - 00010933
WHDL - 00010933
Submitted to the Department of Mathematics and Computer Science in partial fulfillment of the requirements for the degree of Bachelor of ScienceThis is a system for analyzing post-fire imagery to determine wildland fire severity. The system is written for the FireMAP project. It is written in C++ and C using the open source image processing library OpenCV. The system primarily is comprised of a k-Dimensional binary tree, for storing training data, along with a k-Nearest Neighbors algorithm to quickly classify imagery based on the training data. The algorithm utilizes parallel processing to fully utilize the CPU greatly increase the classification speed. Most of the system is written in-house to provide a unique and modifiable algorithm implementation for use by the FireMAP project. This effort is important to the FireMAP project because it provides the ability to automate the severity determination process that would previously take weeks for a Burn Area Emergency Response (BAER) team to do now takes several minutes or several seconds. So, the solution provided by this project is cheaper, faster, requires less manpower, and is overall a safer approach to the issue of burn severity analysis. Several areas of research and work were not included in this project and will require further attention. These areas are object-based analysis of imagery, further optimization and training for the classifier, more robust system testing, and full integration of the classifier into the full FireMAP project.