Spectral and Spatial Pattern Recognition in Digital Imagery
Abstract
The main purpose of this research is to develop a model that can be used to solve combined spectral and spatial pattern recognition problems. The basis of my model is a multiobjective discrete programming model developed by Benabdullah and Wright (B and W). The model will be modified and then tested by solving a real world problem with SPOT multispectral imagery. Several improvements were added to the B and W model, namely standard border length accounting and control over which pixels are selected by the model. The model was also improved to process more than one channel of imagery at a time. The model was successful in locating a user specified target, but this was not possible with all SPOT channel combinations. The Channel 2-3 combination caused the program to abort after 5,000 iterations. Three improvements are still required. An ADBASE formulation would automatically try all the different lambda weights and thus find all the noninferior solutions. Another improvement is to reformulate the problem as a network with side constraints problem. This would ensure integer solutions and quicker results. The final improvement is the creation of an objective function that selects the most representative pixels of a particular land type.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1992
- Accession Number
- ADA258910
Entities
People
- John M. Amrine
Organizations
- Air Force Institute of Technology