A Statistical Analysis of Spatial Classifiers for a Landsat TM Mapping Problem with Incomplete Spatial Coverage
Document Type
Presentation Abstract
Presentation Date
11-2-2000
Abstract
The theme of this talk is a new hybrid of statistical classification rules used for land cover mapping. A land cover type map identifies the dominant vegetation, or land surface types for a region. A Landsat TM land cover map is produced from a base map consisting of as many as 1 million polygons, each of which belongs to an unknown land cover type. Usually, land cover type is assigned to polygons using a classification rule constructed from remotely sensed variables and a training set obtained from ground sampling. Spatial classifiers are classification rules that use spatial information extracted from the training set in addition to remotely sensed variables. Optimally, the distribution of sample locations should cover the map region; however, good coverage often is not possible. With incomplete coverage, accuracy is compromised and estimates of map accuracy are highly suspect. This talk discusses the statistical analysis of spatial classifiers for a Landsat TM mapping problem with grossly poor spatial coverage.
Recommended Citation
Steele, Professor Brian, "A Statistical Analysis of Spatial Classifiers for a Landsat TM Mapping Problem with Incomplete Spatial Coverage" (2000). Colloquia of the Department of Mathematical Sciences. 76.
https://scholarworks.umt.edu/mathcolloquia/76
Additional Details
Thursday, 2 November 2000
4:10 p.m. in Math 109
Coffee/treats at 3:30 p.m. Math 104 (Lounge)