Anomaly Segmentation for High-Resolution Remote Sensing Images Based on Pixel Descriptors

Anomaly segmentation in high spatial resolution (HSR) remote sensing imagery is aimed at segmenting anomaly patterns of the earth deviating from normal patterns, which plays an important role various Earth vision applications. However, it a challenging task due to complex distribution and irregular shapes objects, lack abnormal samples. To tackle these problems, model based on pixel descriptors (ASD) proposed for HSR imagery. Specifically, deep one-class classification introduced feature space with discriminative descriptors. The ASD incorporates data argument generating virtual samples, can force be compact meanwhile diverse avoid collapse problems when only positive samples participated training. In addition, multi-level multi-scale extraction strategy learning low-level semantic information make feature-rich. was validated using four datasets compared recent state-of-the-art models, showing its potential value