The Sample Size Influence in the Accuracy of the Image Classification of the Remote Sensing

Abstract

Landuse/landcover maps produced by classification of remote sensing images incorporate uncertainty. This uncertainty is measured by accuracy indices using reference samples. The size of the reference sample is defined by approximation by a binomial function without the use of a pilot sample. This way the accuracy are not estimated, but fixed a priori. In case of divergency between the estimated and a priori accuracy the error of the sampling will deviate from the expected error. The size using pilot sample (theorically correct procedure) justify when haven´t estimate of accuracy for work area, referent the product remote sensing utility.

Keywords

• accuracy
• reference sample
• image classification
• remote sensing
• landuse/landcover.