State of the art of Landsat classification accuracy assessment.
Methods of assessing Landsat classification accuracy are necessary in order for any classification to be evaluated and improved. Continuous improvement of the ways to assess accuracy is an important key to the further use of Landsat data. Employed methods of accuracy determination and analysis will assist investigators in developing applications of Landsat imagery. The most common mans of expressing accuracy is a percentage based on the number of correctly classified pixels divided Sap the total number tallied from ground truth. This assessment is usually done as an afterthought without much regard for statistical correctness. As a result, there are no standardized methods or standardized terminology for accuracy determination. Problems also arise in failure to differentiate between site specific and non-site specific accuracy. However, other more powerful means of determination and analysis have been tried and better methods are now being developed.
Natural Resources and the Environment
Proceedings of the Sixth Annual Symposium on Machine Processing of Remotely Sensed Data
Congalton, R. G. and R. A. Mead. 1980. State of the art of Landsat classification accuracy assessment. Proceedings of the Sixth Annual Symposium on Machine Processing of Remotely Sensed Data. Purdue University, West Lafayette, IN. p. 337.
Copyright © 1980 IEEE