A Robust Dynamic Classifier Selection Approach for Hyperspectral Images with Imprecise Label Information

Abstract

Supervised hyperspectral image (HSI) classification relies on accurate label information. However, it is not always possible to collect perfectly accurate labels for training samples. This motivates the development of classifiers that are sufficiently robust to some reasonable amounts of errors in data labels. Despite the growing importance of this aspect, it has not been sufficiently studied in the literature yet. In this paper, we analyze the effect of erroneous sample labels on probability distributions of the principal components of HSIs, and provide in this way a statistical analysis of the resulting uncertainty in classifiers. Building on the theory of imprecise probabilities, we develop a novel robust dynamic classifier selection (R-DCS) model for data classification with erroneous labels. Particularly, spectral and spatial features are extracted from HSIs to construct two individual classifiers for the dynamic selection, respectively. The proposed R-DCS model is based on the robustness of the classifiers’ predictions: the extent to which a classifier can be altered without changing its prediction. We provide three possible selection strategies for the proposed model with different computational complexities and apply them on three benchmark data sets. Experimental results demonstrate that the proposed model outperforms the individual classifiers it selects from and is more robust to errors in labels compared to widely adopted approaches.

Exponential negation of a probability distribution

“…Knowledge representation and uncertainty measure are important issues in artificial intelligence [14,27,7,57,31]. A lot of methods have been presented, including intuitionistic fuzzy sets [16], Z-number [19], evidence theory [38], D numbers [25], evidential reasoning [58,24], complex evidence theory [45,46,47], which are applied in classification [28,52,23], information fusion [48,22], medical diagnosis [3,43], fault diagnosis [50], intrusion detection [32], reliability analysis [32], risk analysis and assessment [34,56,33,44,20], and decision making [42,11,12,8].…”

Decision making under measure-based granular uncertainty with intuitionistic fuzzy sets

“…Nowadays, there are a lot of uncertain issues [23][24][25][26]. The real world is very complicated [27][28][29]. The unknown is everywhere, just like no one knows why there will be an outbreak of COVID-19 in early 2020, no one knows when the next earthquake will come, no one knows when the next tsunami will come.…”

A fast evidential approach for stock forecasting

“…Uncertain decision-making is widely used in many ways. [28][29][30][31] Evidence theory supposes the frame of discernment (FOD) which is the definition of a set of hypotheses as follows:…”

“…Uncertain decisions can be made under unstable conditions in a variety of different options. Uncertain decision‐making is widely used in many ways 28–31 …”