Tim Hammond


M.S. Natural Resources Management, University of Alaska Fairbanks,1995

Current Positions:  Senior GIS Analyst, Alaska Fire Service, Bureau of Land Managment

email: thammon@ak.blm.gov

Published Arcticles from M. S. Thesis:

Optimistic Bias in Classification Accuracy Assessment

International Journal of Remote Sensing 17(6):1261-1266.

ABSTRACT

There are many sources of both conservative and optimistic bias in classification accuracy assessment. In this paper, we discuss three sources of optimistic bias: 1)use of training data for accuracy assessment, 2)restriction of reference data sampling to homogeneous areas, and 3)sampling of reference data not independent of training data. The magnitude and direction of bias in classification accuracy estimates depends on the methods used for classification and reference data sampling. However, based on our review of 1994 papers published in three remote sensing journals, we conclude that many studies currently do not report their methods in sufficient detail to enable readers to assess the potential for bias in classification accuracy estimates.


Conservative Bias in Classification Accuracy Assessment

International Journal of Remote Sensing 16(3):581-587.

ABSTRACT

The use of reference grids derived from aerial photography for a pixel- by-pixel comparison with classified images can yield conservative estimates of classification accuracy. Even if the class assignment of each polygon is 100 percent correct, and there is no change in cover type due to temporal differences between the reference data and the classified image, conservative bias in estimates of classification accuracy are still possible. In this paper, we discuss two major sources of this bias: 1)positional errors, and 2) difference between polygon minimum mapping unit area and pixel size of the classified image.
 


Last modified 15 August 1997

 email: thammon@ak.blm.gov