Active learning for interactive satellite image change detection

TL;DR

This paper presents a new interactive active learning algorithm for satellite image change detection that efficiently selects the most informative image pairs for annotation, improving detection accuracy after natural hazards.

Contribution

The paper introduces a novel active learning framework that models sample relevance probability by optimizing for representativity, diversity, and ambiguity, tailored for satellite image change detection.

Findings

Outperforms existing methods in detecting changes after natural hazards

Efficiently selects highly relevant samples for annotation

Improves change detection accuracy with fewer labeled samples

Abstract

We introduce in this paper a novel active learning algorithm for satellite image change detection. The proposed solution is interactive and based on a question and answer model, which asks an oracle (annotator) the most informative questions about the relevance of sampled satellite image pairs, and according to the oracle's responses, updates a decision function iteratively. We investigate a novel framework which models the probability that samples are relevant; this probability is obtained by minimizing an objective function capturing representativity, diversity and ambiguity. Only data with a high probability according to these criteria are selected and displayed to the oracle for further annotation. Extensive experiments on the task of satellite image change detection after natural hazards (namely tornadoes) show the relevance of the proposed method against the related work.