Documents in the NTIS Technical Reports collection are the results of federally funded research. They are directly submitted to or collected by NTIS from Federal agencies for permanent accessibility to industry, academia and the public.  Before purchasing from NTIS, you may want to check for free access from (1) the issuing organization's website; (2) the U.S. Government Printing Office's Federal Digital System website; (3) the federal government Internet portal; or (4) a web search conducted using a commercial search engine such as
Accession Number PB2013-100702
Title Machine-Learning Approach to Change Detection Using Multi-Scale Imagery.
Publication Date 2012
Media Count 10p
Personal Author B. Maurizi C. Fischer J. Suero L. M. Levien P. Roffers X. Huang
Abstract The USDA Forest Service and the California Department of Forestry and Fire Protection are collaborating on a statewide change detection program to identify landcover change across all ownerships within five-year time periods. This program uses Landsat Thematic Mapper satellite imagery to derive landcover change and aid in assessing its cause. Landscape changes are initially detected using a multi-temporal Kauth-Thomas transform. Unsupervised classification of the transformed imagery creates a preliminary landscape-level change map portraying change classes with multiple levels of vegetation increase, decrease, and no change. Using a stratified random sampling scheme, this preliminary change map facilitates selection of field sites for collecting vegetation canopy cover measurements. Ground truth for the classifier is obtained by estimating canopy cover change over the 5-year timeframe using color-IR digital photos, digital orthophoto quads, and aerial photography. Canopy cover estimates from the second date of photography are calibrated using transect measurements of canopy cover from a sample of field sites. Attributes such as species and vegetative cover are also noted. A machine learning classifier approach is then employed. The classifier uses an inductive learning algorithm to generate production rules from training data, including the transformed change data and other ancillary data layers. The resultant knowledge base is then used by an expert classifier to produce classes of crown closure change. Approximately half of the field.
Keywords Detection
Image processing
Monitor changes
Remote sensing
Satellite imagery

Source Agency Forest Service Engineering Staff Washington DC
NTIS Subject Category 84G - Unmanned Spacecraft
63 - Detection & Countermeasures
46C - Optics & Lasers
Corporate Author California Dept. of Forestry and Fire Protection, Sacramento.
Document Type Technical report
Title Note N/A
NTIS Issue Number 1307
Contract Number N/A

Science and Technology Highlights

See a sampling of the latest scientific, technical and engineering information from NTIS in the NTIS Technical Reports Newsletter

Acrobat Reader Mobile    Acrobat Reader