基于局部对比度和多向梯度的高光谱异常检测

武丽; 徐星臣; 王一安; 任佳红; 张嘉嘉; 赵东; 王新蕾

基于局部对比度和多向梯度的高光谱异常检测

武丽^{1, 2,},
徐星臣^{1, 2},
王一安³,
任佳红¹,
张嘉嘉⁴,
赵东^{1, 2},
王新蕾^{1, 2, ,}

1.
南京信息工程大学电子与信息工程学院, 江苏南京 210044
2.
无锡学院电子信息工程学院, 江苏无锡 214105
3.
西安石油大学电子工程学院, 陕西西安 710071
4.
西安电子科技大学物理学院, 陕西西安 710071

基金项目:

国家自然科学基金 2105258

江苏省自然科学基金 BK20210064

无锡市创新创业资金“太湖之光”科技攻关计划（基础研究）项目 K20221046

无锡学院人才启动基金 2021r007

详细信息

作者简介:
武丽（1983-），女，江苏东台人，硕士，副教授，硕导，主要研究方向为信号处理、模式识别，E-mail: wuli@cwxu.edu.cn

通讯作者:
王新蕾（1980-），女，山东潍坊人，博士，讲师，硕导，主要研究方向为机器学习、模式识别，E-mail: wangxinlei@cwxu.edu.cn

中图分类号: TP751.1
计量
- 文章访问数: 39
- HTML全文浏览量: 1
- PDF下载量: 5
出版历程
- 收稿日期: 2024-04-27
- 修回日期: 2024-05-29
- 网络出版日期: 2025-05-27
- 刊出日期: 2025-05-19

Hyperspectral Anomaly Detection Based on Local Contrast and Multidirectional Gradients

WU Li^{1, 2,},
XU Xingchen^{1, 2},
WANG Yian³,
REN Jiahong¹,
ZHANG Jiajia⁴,
ZHAO Dong^{1, 2},
WANG Xinlei^{1, 2, ,}

1.
School of Electronics and Information Engineering, Nanjing University of Information Science and Technology, Nanjing 210044, China
2.
School of Electronics and Information Engineering, Wuxi University, Wuxi 214105, China
3.
School of Electronics and Information Engineering, Xi'an Shiyou University, Xi'an 710071, China
4.
School of Physics, Xi'an University of Electronic Science and Technology, Xi'an 710071, China

摘要

摘要:
为了充分利用高光谱图像的空间和光谱信息，并抑制图像中的噪声，提出了一种基于局部对比度和多向梯度的高光谱异常检测方法。首先，为利用局部光谱信息，提出了一种局部对比度策略，通过计算目标与背景之间的亮度差异，获得光谱检测得分图。然后，为了降低计算的复杂性，引入了一种光谱融合降维技术对高光谱图像进行处理。此外，提出了一种局部多向梯度特征方法，旨在减少图像噪声和保留局部细节特征，生成多向梯度检测得分图。最后，通过融合两张得分图，得到最终的异常结果图。实验结果表明，在4个经典数据集上本文方法能够成功展示异常目标，并且相较于其他7种方法，其检测精度更高、虚警率更低。
- 高光谱图像 /
- 异常检测 /
- 局部对比度 /
- 光谱融合降维 /
- 多向梯度特征
Abstract:
To fully utilize the spatial and spectral information of hyperspectral images and suppress image noise, a hyperspectral anomaly detection method based on local contrast and multidirectional gradient analysis is proposed. First, to leverage local spectral information, a local contrast strategy is introduced, generating a spectral detection score map based on the brightness difference between the target and the background. Then, to reduce computational complexity, a spectral fusion-based dimensionality reduction technique is proposed to process hyperspectral images. In addition, a local multidirectional gradient feature method is proposed to reduce image noise, retain local detail features, and generate a multidirectional gradient detection score map. Finally, the anomaly result map is obtained by fusing the spectral and gradient-based score graphs. Experimental results on four classical datasets demonstrate that the proposed method can successfully display abnormal targets in the result graph, achieving higher detection accuracy and lower false alarm rates compared to seven existing methods.
- hyperspectral image /
- anomaly detection /
- local contrast /
- spectral fusion dimensionality reduction /
- multi-directional gradient features

HTML全文

图 1 HLC-MDG方法的总流程图

Figure 1. Overall flow chart of the HLC-MDG method

下载: 全尺寸图片幻灯片

图 2 实验数据集的伪彩图和真实图

Figure 2. False color images and ground-truth images of the experimental data set

下载: 全尺寸图片幻灯片

图 3 不同参数对HLC-MDG方法性能的影响

Figure 3. Effect of different parameters on the performance of HLC-MDG method

下载: 全尺寸图片幻灯片

图 4 内外窗尺寸对HLC-MDG方法性能的影响

Figure 4. Effect of inner and outer window size on the performance of HLC-MDG method

下载: 全尺寸图片幻灯片

图 5 不同算法对4个数据集的检测结果

Figure 5. Detection results of different algorithms on four data sets

下载: 全尺寸图片幻灯片

图 6 四个数据集上不同算法的ROC曲线（P_D, P_F）

Figure 6. ROC curves of different algorithms on four data sets (P_D, P_F)

下载: 全尺寸图片幻灯片

图 7 四个数据集上不同算法的ROC曲线（P_D, τ）

Figure 7. ROC curves of different algorithms on four data sets (P_D, τ)

下载: 全尺寸图片幻灯片

图 8 四个数据集上不同算法的ROC曲线(P_F, τ)

Figure 8. ROC curves of different algorithms on four data sets (P_F, τ)

下载: 全尺寸图片幻灯片

表 1 四个数据集上不同算法的AUC_{(D, F)}值

Table 1 AUC_{(D, F)} values of different algorithms on four datasets

Dataset	Method
Dataset	GRX	LRX	LRASR	LSMAD	CRD	GTVLRR	AHMID	HLSC-MDGF
Gulfport	0.9525	0.9810	0.9563	0.9864	0.8862	0.9843	0.9857	0.9960
HYDICE	0.9931	0.9850	0.9788	0.9902	0.9811	0.9817	0.9770	0.9951
San Diego Ⅰ	0.9411	0.9748	0.9859	0.9830	0.9298	0.9943	0.9928	0.9991
San Diego Ⅱ	0.9832	0.9807	0.9661	0.9870	0.9788	0.9957	0.9845	0.9912

下载: 导出CSV

表 2 四个数据集上不同算法的AUC_{(D, τ)}值

Table 2 AUC_{(D, τ)} values of different algorithms on four datasets

Dataset	Method
Dataset	GRX	LRX	LRASR	LSMAD	CRD	GTVLRR	AHMID	HLSC-MDGF
Gulfport	0.0736	0.1447	0.2174	0.3797	0.0093	0.4438	0.2436	0.3860
HYDICE	0.2189	0.3031	0.5314	0.2337	0.3384	0.2166	0.3987	0.4143
San Diego Ⅰ	0.1017	0.1670	0.5425	0.2141	0.0039	0.4536	0.4114	0.6080
San Diego Ⅱ	0.2889	0.0316	0.3007	0.2136	0.2333	0.4717	0.0657	0.2938

下载: 导出CSV

表 3 四个数据集上不同算法的AUC_{(F, τ)}值

Table 3 AUC_{(F, τ)} values of different algorithms on four datasets

Dataset	Method
Dataset	GRX	LRX	LRASR	LSMAD	CRD	GTVLRR	AHMID	HLSC-MDGF
Gulfport	0.0248	0.0076	0.1113	0.0587	0.0063	0.0823	0.0317	0.0032
HYDICE	0.0331	0.0032	0.0466	0.0195	0.0420	0.0471	0.0106	0.0024
San Diego Ⅰ	0.0470	0.0338	0.0627	0.0404	0.0023	0.1263	0.0212	0.0037
San Diego Ⅱ	0.0714	0.0046	0.0855	0.0281	0.1110	0.0307	0.0059	0.0032

下载: 导出CSV

表 4 HLC-MDG中有无LSC的AUC_{(D, F)}值

Table 4 AUC_{(D, F)} values of HLC-MDG with and without LSC

Method	AUC_{(D, F)}
Method	Gulfport	HYDICE	San Diego Ⅰ	San Diego Ⅱ
+LSC	0.9960	0.9951	0.9991	0.9912
-LSC	0.9688	0.9743	0.9803	0.9725

下载: 导出CSV

表 5 SFRD和两种常用降维方法的AUC_{(D, F)}值

Table 5 AUC_{(D, F)} values of SFRD and two commonly used dimensionality reduction methods

Method	AUC_{(D, F)}
Method	Gulfport	HYDICE	San Diego Ⅰ	San Diego Ⅱ
SFRD	0.9960	0.9951	0.9991	0.9912
PCA	0.9885	0.9847	0.9922	0.9895
FS	0.9879	0.9920	0.9930	0.9866

下载: 导出CSV

表 6 HLC-MDG中有无LMGC的AUC_{(D, F)}值

Table 6 AUC_{(D, F)} values of HLC-MDG with and without LMGC

Method	AUC_{(D, F)}
Method	Gulfport	HYDICE	San Diego Ⅰ	San Diego Ⅱ
+LMGC	0.9960	0.9951	0.9991	0.9912
-LMGC	0.9839	0.9843	0.9903	0.9852

下载: 导出CSV

参考文献(29)

[1]	孙帮勇, 赵哲, 胡炳樑, 等. 基于3D卷积自编解码器和低秩表示的高光谱异常检测[J]. 光子学报, 2021, 50(4): 262-274. SUN Bangyong, ZHAO Zhe, HU Bingliang, et al. Hyperspectral anomaly detection based on 3D convolutional autoencoder and low rank representation[J]. Acta Photonica Sinica, 2021, 50(4): 262-274.
[2]	XU Y, ZHANG L, DU B, et al. Hyperspectral anomaly detection based on machine learning: an overview[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15: 3351-3364. DOI: 10.1109/JSTARS.2022.3167830
[3]	李茗欣, 黄远程, 竞霞, 等. 融合视觉注意机制的高光谱RX异常检测算法[J]. 红外技术, 2023, 45(4): 402-409. http://hwjs.nvir.cn/article/id/999a2986-db0e-4fe0-a22e-95fcc5a1b699 LI Mingxin, HUANG Yuancheng, JING Xia, et al. Hyperspectral RX anomaly detection algorithm with visual attention mechanism[J]. Infrared Technology, 2023, 45(4): 402-409. http://hwjs.nvir.cn/article/id/999a2986-db0e-4fe0-a22e-95fcc5a1b699
[4]	向英杰, 张俭峰, 杨桄, 等. 基于混合噪声估计的高光谱图像异常检测方法[J]. 红外技术, 2017, 39(8): 734-739. http://hwjs.nvir.cn/article/id/hwjs201708011 XIANG Yingjie, ZHANG Jianfeng, YANG Guang, et al. A mixed-noise estimation-based anomaly detection method for hyperspectral image[J]. Infrared Technology, 2017, 39(8): 734-739. http://hwjs.nvir.cn/article/id/hwjs201708011
[5]	SONG X, LING Z, WU L, et al. Hyperspectral image anomaly detection based on background reconstruction[J]. Journal of System Simulation, 2020, 32(7): 1287-1293.
[6]	XIANG P, SONG J, QIN H, et al. Visual attention and background subtraction with adaptive weight for hyperspectral anomaly detection[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 14: 2270-2283. DOI: 10.1109/JSTARS.2021.3052968
[7]	ZHAO D, Asano Y, GU L, et al. City-scale distance sensing via bispectral light extinction in bad weather[J]. Remote Sensing, 2020, 12(9): 1401. DOI: 10.3390/rs12091401
[8]	ZHAO D, ZHOU L, LI Y, et al. Visibility estimation via near-infrared bispectral real-time imaging in bad weather[J]. Infrared Physics & Technology, 2024, 136: 105008.
[9]	ZHANG J, XU X, YAN W, et al. Hyperspectral anomaly detection based on local contrast estimation and sub-block background estimation[J]. Infrared Physics & Technology, 2023, 135: 104966.
[10]	Reed I S, YU X. Adaptive multiple-band CFAR detection of an optical pattern with unknown spectral distribution[J]. IEEE Transactions on Acoustics, Speech, and Signal Processing, 1990, 38(10): 1760-1770. DOI: 10.1109/29.60107
[11]	Molero J M, Garzon E M, Garcia I, et al. Analysis and optimizations of global and local versions of the RX algorithm for anomaly detection in hyperspectral data[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2013, 6(2): 801-814. DOI: 10.1109/JSTARS.2013.2238609
[12]	GUO Q, ZHANG B, RAN Q, et al. Weighted-RXD and linear filter-based RXD: Improving background statistics estimation for anomaly detection in hyperspectral imagery[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 7(6): 2351-2366. DOI: 10.1109/JSTARS.2014.2302446
[13]	Kwon H, Nasrabadi N M. Kernel RX-algorithm: A nonlinear anomaly detector for hyperspectral imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2005, 43(2): 388-397. DOI: 10.1109/TGRS.2004.841487
[14]	LIU G, LIN Z, YAN S, et al. Robust recovery of subspace structures by low-rank representation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 35(1): 171-184.
[15]	LI S, WANG W, QI H, et al. Low-rank tensor decomposition based anomaly detection for hyperspectral imagery[C]//2015 IEEE International Conference on Image Processing (ICIP). IEEE, 2015: 4525-4529.
[16]	LI W, DU Q. Collaborative representation for hyperspectral anomaly detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2014, 53(3): 1463-1474.
[17]	CHEN Y, Nasrabadi N M, Tran T D. Sparse representation for target detection in hyperspectral imagery[J]. IEEE Journal of Selected Topics in Signal Processing, 2011, 5(3): 629-640. DOI: 10.1109/JSTSP.2011.2113170
[18]	WEI J, ZHANG J, XU Y, et al. Hyperspectral anomaly detection based on graph regularized variational autoencoder[J]. IEEE Geosci. Remote Sens. Lett. , 2022, 19: 1-5.
[19]	Banks M S, Read J C A, Allison R S, et al. Stereoscopy and the human visual system[J]. SMPTE Motion Imaging Journal, 2012, 121(4): 24-43. DOI: 10.5594/j18173
[20]	XIE W, FAN S, QU J, et al. Spectral distribution-aware estimation network for hyperspectral anomaly detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 1-12.
[21]	XU Y, WU Z, LI J, et al. Anomaly detection in hyperspectral images based on low-rank and sparse representation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 54(4): 1990-2000.
[22]	ZHANG Y, DU B, ZHANG L, et al. A low-rank and sparse matrix decomposition-based Mahalanobis distance method for hyperspectral anomaly detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 54(3): 1376-1389.
[23]	CHENG T, WANG B. Graph and total variation regularized low-rank representation for hyperspectral anomaly detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 58(1): 391-406.
[24]	GUO T, HE L, LUO F, et al. Anomaly detection of hyperspectral image with hierarchical anti-noise mutual-incoherence-induced low-rank representation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 1-13.
[25]	Tu B, Li N, Liao Z, et al. Hyperspectral anomaly detection via spatial density background purification[J]. Remote Sensing, 2019, 11(22): 2618. DOI: 10.3390/rs11222618
[26]	Song S, Zhou H, Yang Y, et al. Hyperspectral anomaly detection via convolutional neural network and low rank with density-based clustering[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019, 12(9): 3637-3649. DOI: 10.1109/JSTARS.2019.2926130
[27]	Xiang P, Ali S, Jung S K, et al. Hyperspectral anomaly detection with guided autoencoder[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-18.
[28]	Arisoy S, Nasrabadi N M, Kayabol K. Unsupervised pixel-wise hyperspectral anomaly detection via autoencoding adversarial networks[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 19: 1-5.
[29]	钱晓亮, 曾银凤, 林生, 等. 融合自适应窗口显著性检测和改进超像素分割的高光谱异常检测[J]. 遥感学报, 2023, 27(12): 2748-2761. QIAN Xiaoliang, ZENG Yinfeng, LIN Sheng, et al. Hyperspectral anomaly detection via combining adaptive window saliency detection and improved superpixel segmentation[J]. National Remote Sensing Bulletin, 2023, 27(12): 2748-2761.

施引文献(26)

期刊类型引用(13)

1.	王振，刘磊. 融合彩色模型空间的电力设备红外图像增强. 红外技术. 2024(02): 225-232 . 本站查看
2.	王奎，黄福珍. 基于残差融合的改进Retinex图像增强算法. 计算机应用与软件. 2024(04): 205-211+218 . 百度学术
3.	马瑞. 基于自适应遗传算法的红外子热像图模糊增强方法. 湖南文理学院学报(自然科学版). 2024(03): 30-36 . 百度学术
4.	万卫国，李纯洁，陈孟秋. 大电流供电系统智能故障预警. 船电技术. 2023(05): 29-31 . 百度学术
5.	何智博，曾祥进，邓晨，宋彭彭. 基于局部熵-局部对比度和双区域直方图均衡化的红外图像增强. 红外技术. 2023(06): 598-604 . 本站查看
6.	肖鹏，王红茹. 一种用于局部低照度水下图像的自适应增强方法. 激光杂志. 2022(04): 114-119 . 百度学术
7.	张丽娟，朱会龙. 基于光照补偿的产品包装外观图像自适应增强方法. 激光杂志. 2022(04): 184-188 . 百度学术
8.	迟明伟. 基于电网热红外图像增强算法质量评价. 科技创新与应用. 2022(13): 83-87 . 百度学术
9.	李凌杰，陈菲菲. 基于改进直方图的红外图像增强方法. 航空兵器. 2022(02): 101-105 . 百度学术
10.	朱家乙，杨宏双，何伟，王伟男，沙怡中，黄晓江，许桢杰. 一种基于区域分割的直方图均衡算法. 红外技术. 2022(06): 587-592 . 本站查看
11.	蔡美芳，万里勇. 基于自适应Gamma校正的红外图像增强算法. 光学技术. 2022(04): 486-491 . 百度学术
12.	李少荣. 基于改进直方图均衡化的红外图像增强技术的研究. 工业控制计算机. 2022(12): 52-53+56 . 百度学术
13.	马琼，李通，赵巨峰，崔光茫. 使用局部视觉显著分析的红外图像增强. 光学技术. 2021(05): 601-607 . 百度学术