论文 | Yubo Ma, Siyu Cai, Jie Zhou:Adaptive Reference-Related Graph Embedding for Hyperspectral Anomaly Detection

本文（Adaptive Reference-Related Graph  Embedding for Hyperspectral Anomaly Detection）原载 IEEE Transactions on  Geoscience and Remote Sensing，由四川大学周杰教授等科研人员创作，系四川大学智慧法治超前部署学科系列学术成果。后续会持续分享四川大学智慧法治超前部署学科系列学术成果，欢迎大家阅读。

Abstract:Graph embedding (GE) provides an effective way to reveal the  intrinsic feature of high-dimensional data on the foundation of preserving  topological properties. Under the framework of GE, the hyperspectral image can be  represented by a weighted graph, where pixels and similarities among them are  treated as vertices and edge weights, respectively. In this article, an adaptive  reference-related GE (ARGE) method is proposed to efficaciously obtain the  low-dimensional feature and improve computational efficiency. The ARGE method  is composed of two primary processes. The key to connecting these two processes is  the reference vertices set, which is the abstraction of graph topological features. First,  the reference vertices are adaptively selected through a three-step adaptive reference  set selection (ARSS) algorithm. Second, the original high-dimensional graph is  embedded as a low-dimensional graph through preserving the reference-related  structure. Specifically, the pairwise similarities between vertices and reference  vertices are preserved in embedding space. In addition, a new hybrid dissimilarity  measure of Rao distance and spectral information divergence (RD-SID) is designed  to depict the spectral difference between pixels. To evaluate the effectiveness of the  proposed method, the obtained low-dimensional feature is fed into the anomaly  detector to detect anomalous pixels. The experimental results on five real and one  synthetic hyperspectral datasets demonstrate the superiority of the proposed ARGE  method over the compared feature extraction methods

Yubo Ma, Siyu Cai, Jie Zhou. Adaptive Reference-Related Graph  Embedding for Hyperspectral Anomaly Detection, IEEE Transactions on  Geoscience and Remote Sensing, Vol.61, Article 5504514, February 2023.（论文下载）