A survey of active learning algorithms for supervised remote sensing image classification

TL;DR

This survey reviews and tests various active learning algorithms for remote sensing image classification, providing guidelines for selecting effective methods in complex scenarios like hyperspectral and high-resolution images.

Contribution

It offers a comprehensive review and empirical testing of main active learning algorithms tailored for remote sensing, including recent advances and practical guidelines.

Findings

Committee-based algorithms perform well in high-dimensional data

Uncertainty sampling improves labeling efficiency

Guidelines assist new users in selecting suitable algorithms

Abstract

Defining an efficient training set is one of the most delicate phases for the success of remote sensing image classification routines. The complexity of the problem, the limited temporal and financial resources, as well as the high intraclass variance can make an algorithm fail if it is trained with a suboptimal dataset. Active learning aims at building efficient training sets by iteratively improving the model performance through sampling. A user-defined heuristic ranks the unlabeled pixels according to a function of the uncertainty of their class membership and then the user is asked to provide labels for the most uncertain pixels. This paper reviews and tests the main families of active learning algorithms: committee, large margin and posterior probability-based. For each of them, the most recent advances in the remote sensing community are discussed and some heuristics are detailed and tested. Several challenging remote sensing scenarios are considered, including very high spatial resolution and hyperspectral image classification. Finally, guidelines for choosing the good architecture are provided for new and/or unexperienced user.