Detection of Large-Scale and Anomalous Changes

We survey algorithms and methodologies for detecting and delineating changes of interest in remote sensing imagery. We consider both broad salient changes and rare anomalous changes, and we describe strategies for exploiting imagery containing these changes. The perennial challenge in change detection is in translating the application-dependent concept of an “interesting change” to a mathematical framework; as such, the mathematical approaches for detecting these types of changes can be quite different. In large-scale change detection (LSCD), the goal is to identify changes that have broadly occurred in the scene. The paradigm for anomalous change detection (ACD), which is grounded in concepts from anomaly detection, seeks to identify changes that are different from how everything else might have changed. This borrows from the classic anomaly detection framework, which attempts to characterize that which is “typical” and then uses that to identify deviations from what is expected or common. This chapter provides an overview of change detection, including a discussion of LSCD and ACD approaches, operational considerations, relevant datasets for testing the various algorithms, and some illustrative results.