Abstract
Abstract. A majority of studies involving remote sensing LULC classification conducted classification accuracy assessment without consideration of the training data uncertainty. In this study we present new concepts of LULC classification accuracies, namely the training-sample-based global accuracy and the classifier global accuracy, and a general expression of different measures of classification accuracy in terms of the sample dataset for classifier training and the sample dataset for evaluation of classification results. Through stochastic simulation of a two-feature and two-class case, we demonstrate that the training-sample confusion matrix should replace the commonly adopted reference-sample confusion matrix for evaluation of LULC classification results. We then propose a bootstrap-simulation approach for establishing 95% confidence intervals of classifier global accuracies.
Highlights
1.1 General InstructionsWhen conducting a supervised LULC classification using remote sensing images, a set of multi-class ground-truth training samples is collected and used to establish classification rules and multi-class boundaries in the feature space
Different measures of classification accuracy of the reference samples are summarized in a confusion matrix for assessment of the classification accuracies and performance of the LULC classification
A critical assumption of the classification accuracy analysis is that the confusion matrix is truly representative of the classification results of the entire study area
Summary
1.1 General InstructionsWhen conducting a supervised LULC classification using remote sensing images, a set of multi-class ground-truth training samples is collected and used to establish classification rules and multi-class boundaries in the feature space. Class-specific producer’s and user’s accuracy summarized in a confusion matrix can be considered as sample accuracy and are only estimates of the true, yet unknown, global accuracy (or population accuracy) concerning the entire study area (Hay, 1988; Stehman and Czaplewski, 1998). These accuracies or errors are inherently associated with uncertainties due to variability or uncertainty in selection of training and reference samples (Weber and Langille, 2007)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have