Forest managers are in constant need of accurate, up-to-date resource information. A first attempt towards an operational, quantitative, remote-sensing-based change detection system is described. The change information derived from this system can then be used either to "flag" those areas that require additional detailed investigation, or to monitor conditions to determine if changes occur as expected. The digital change detection system described is based on standardized differences of Kauth-Thomas transformations. Minimum-distance, maximum-likelihood, and Mahalanobis-distance classifiers were tested with field data and compared. The maximum-likelihood and Mahalanobis-distance classifiers produced the more accurate results. They were able to detect small amounts of change resulting from forest thinnings, which are the most difficult to discriminate. Overall results of this work demonstrated the high potential value of an operational, digital, quantitative change detection system to support forest management decisions across large geographic extents.
|Number of pages
|Photogrammetric Engineering and Remote Sensing
|Published - 2001