Graph Regularized Low-Rank and Collaborative Representation for Hyperspectral Anomaly Detection

Qi Wu; Yanguo Fan; Bowen Fan; Dingfeng Yu

doi:10.3788/LOP202259.1228003

Laser & Optoelectronics Progress, Volume. 59, Issue 12, 1228003(2022)

Graph Regularized Low-Rank and Collaborative Representation for Hyperspectral Anomaly Detection

Qi Wu^1、*, Yanguo Fan¹, Bowen Fan², and Dingfeng Yu³

¹College of Oceanography and Space Informatics, China University of Petroleum (East China), Qingdao 266580, Shandong , China

²College of Underwater Acoustic Engineering, Harbin Engineering University, Harbin 150001, Heilongjiang , China

³Institute of Oceanographic Instrumentation, Qilu University of Technology (Shandong Academy of Sciences), Qingdao 266061, Shandong , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(13)

Fig. 1. Scatter plot of all the pixels in HYDICE data set

Download full size

Fig. 2. Flow chart of hyperspectral anomaly detection based on graph regularized low-rank and collaborative representation

Download full size

Fig. 3. Analysis of endmember number on three datasets. (a) Simulated dataset; (b) HYDICE dataset; (c) Gulfport dataset

Download full size

Fig. 4. Hyperspectral synthetic dataset. (a) Original image of the study area; (b) false-color image of simulated dataset; (c) ground-truth map

Download full size

Fig. 5. HYDICE dataset. (a) Whole image scene; (b) false-color image of the selected region; (c) ground-truth map

Download full size

Fig. 6. Gulfport dataset. (a) False-color image; (b) ground-truth map

Download full size

Fig. 7. Detection accuracy of GLRCRD on the simulated dataset under different parameters. (a) $λ$ variation; (b) $γ$ variation; (c) $β$ variation; (d) $k_{n}$ variation; (e) $σ$ variation

Download full size

Fig. 8. Detection results obtained by six algorithms on the simulated dataset. (a) RX; (b) CRD; (c) LRASR; (d) LSMAD;(e) LRCRD; (f) GLRCRD

Download full size

Fig. 9. Detection results obtained by six slgorithms on the HYDICE dataset. (a) RX; (b) CRD; (c) LRASR;(d) LSMAD; (e) LRCRD; (f) GLRCRD

Download full size

Fig. 10. Detection results obtained by six algorithms on the Gulfport dataset. (a) RX; (b) CRD; (c) LRASR; (d) LSMAD; (e) LRCRD; (f) GLRCRD

Download full size

Fig. 11. ROC curves obtained by six algorithms. (a) Simulated dataset; (b) HYDICE dataset; (c) Gulfport dataset

Download full size

Table 1. AUC values obtained by different anomaly detection algorithms
View table
Table 1. AUC values obtained by different anomaly detection algorithms
Dataset RX CRD LRASR LSMAD LRCRD GLRCRD
Simulated dataset 0.8074 0.8592 0.9299 0.9505 0.9586 0.9709
HYDICE 0.9857 0.9506 0.9765 0.9905 0.9944 0.9970
Gulfport 0.9526 0.9767 0.9532 0.9861 0.9812 0.9910

Table 2. Computation time of different anomaly detection algorithms
View table
Table 2. Computation time of different anomaly detection algorithms
Dataset RX CRD LRASR LSMAD LRCRD GLRCRD
Simulated dataset 0.2254 14.053 86.469 21.216 148.109 350.27
HYDICE 0.1592 7.6372 47.618 9.4717 81.724 233.87
Gulfport 0.1758 9.4595 55.811 13.162 97.930 260.89

Tools

Get Citation

Copy Citation Text

Qi Wu, Yanguo Fan, Bowen Fan, Dingfeng Yu. Graph Regularized Low-Rank and Collaborative Representation for Hyperspectral Anomaly Detection[J]. Laser & Optoelectronics Progress, 2022, 59(12): 1228003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites