3-D Auxiliary Classifier GAN for Hyperspectral Anomaly Detection via Weakly Supervised Learning

• Published in: IEEE geoscience and remote sensing letters 2022, Vol.19, p.1-5
• Main Authors: Luo, Huanlin, Zhu, Haowen, Liu, Shengyang, Liu, Yichuan, Zhu, Xinzhong, Lai, Jinmei
• Format: Article
• Language: English
• Subjects: Algorithms; Anomalies; Anomaly detection (AD); Artificial neural networks; auxiliary classifier generative adversarial network (AC-GAN); Classification algorithms; Classifiers; Detection; Generative adversarial networks; hyperspectral images (HSIs); Hyperspectral imaging; Methods; Network design; Neural networks; Remote observing; Remote sensing; Solid modeling; Spatial discrimination; Spatial resolution; Supervised learning; Tensors; Three-dimensional displays; Training; weakly supervised learning (WSL)
• ISSN: 1545-598X; 1558-0571
• DOI: 10.1109/LGRS.2022.3175836

Hyperspectral anomaly detection (AD) is important in Earth observation and remote sensing. However, the low spatial resolution of hyperspectral images (HSIs), insufficient samples, and lack of prior information limit the detection accuracy. To solve these problems, in this letter, we propose an auxiliary classifier generative adversarial network model based on a 3-D convolutional neural network named 3-D Auxiliary Classifier generative adversarial network (AC-GAN). First, the model is based on a 3-D convolutional neural network design, with 3-D tensors as samples. The network maintains valuable image spatial spectrum joint features to achieve good detection results. It can also generate sufficient samples to achieve dataset augmentation, solving the overfitting problem in GAN training. Second, we train the model with a weakly supervised method. The label of the samples is obtained through the coarse scanning method. Then, the AC-GAN is trained with the bootstrapping method to mitigate the impact of noise labels. The experimental results show that our proposed algorithm outperforms state-of-the-art AD algorithms.