Anomaly Detection in Aerial Imagery Using Color and Texture Features

Abstract

The detection of anomalous regions in digital images can be used in many applications, such as security, search and rescue operations, hazard identification and industrial inspection. In this work, we present an anomaly detection method based on color and texture features applied to a non-linear one-class classifier, and show that it provides excellent results, even when compared to a two-class classifier. Our approach is lightweight and aimed at its implementation on an onboard computer for an Unmanned Aerial Vehicle.