偏振光谱多维信息的重构融合算法
A Multidimensional Information Fusion Algorithm for Polarization Spectrum Reconstruction Based on Nonsubsampled Contourlet Transform
查看参考文献16篇
|
文摘
|
针对传统光学手段难以实现复杂背景下光谱伪装目标的准确识别,同时,常规的数据融合方法易导致图像信息丢失的缺点,提出了一种基于非下采样轮廓波变换的偏振光谱多维信息融合方法。该方法结合自研的新型偏振光谱多维信息探测仪器,根据其获取的目标空间、光谱、偏振等多维信息,设计了多维信息重构算法流程,提取了偏振态基础数据斯托克斯参量以及偏振度和偏振角,利用NSCT对基础偏振参量进行图像融合,提升图像的信息含量以提高伪装物的识别准确率。先对具有相同边缘信息的图像Q和U采用NSCT分解,低通子带取均值,高通子带取最大值进行初步融合,获得偏振特征S,最后对偏振特征S、强度图像I以及偏振度DoLP进行NSCT分解,对分解所得低通子带进行区域能量加权融合;对高通子带,根据偏振特征图像具有灰度值小,受光照影响大等特点,采用LBP特征进行加权融合。同时,本方法与四类融合方法进行对比,据信息熵、标准差、平均梯度、对比度以及峰值信噪比五项指标对融合结果进行客观评价,并结合普通图像,偏振融合图像,偏振高光谱图像对目标识别精度进行对比。融合后的图像信息熵为6.998 6,标准差为45.599 8,平均梯度为19.808 6,与原始强度相比,提升分别为5.1%,14.04%,7.3%,在四类融合方法中排在首位。表明本文所提出的方法有效实现了偏振基础特征融合,提升了人造目标和自然背景的差异。同时融合后的偏振高光谱图像对于目标的识别准确率达到0.986 2,较单一强度图像目标识别准确率提升了21%。实验结果表明,提出的方法能有效融合目标强度信息以及偏振信息,提升图像对比度和可读性,同时融合后的图像在目标识别准确度上有了较大的提升,有效降低了传统光谱手段对伪装目标识别的虚警率,为新概念光谱伪装揭露提供了一种新型有效的手段,具有非常大的应用潜力和应用价值。 |
|
其他语种文摘
|
This paper proposes a polarization spectral multidimensional information fusion method based on nonsubsampled contourlet transform to address the shortcomings of traditional optical methods that make it difficult to identify camouflaged spectral targets in complex backgrounds and the common fusion methods that tend to lead to image information loss.The multidimensional information reconstruction algorithm was designed based on the acquired multidimensional information such as target space,spectrum and polarization,and the basic data of polarization state including Stokes parameters as well as the degree of polarization and angle of polarization were extracted.NSCT is used to fuse the basic polarization parameters to improve the image’s information content and improve the camouflage’s recognition accuracy.The images Q and U with the same edge information are first decomposed using NSCT.Regional energy-weighted fusion is performed for the decomposed low-pass subbands; for the high-pass sub-bands,LBP features are used for weighted fusion according to the characteristics of polarization features,such as small gray value and high influence by illumination.At the same time,the proposed method is compared with four types of fusion methods,and the fusion results are evaluated objectively according to five indicators:information entropy, standard deviation,mean gradient,contrast and peak signal-to-noise ratio,and the target recognition accuracy is compared with plain images,polarized fused images and polarized hyper-spectral images.The information entropy of the fused image is 6.998 6, the standard deviation is 45.599 8,and the average gradient is 19.808 6.Compared with the original intensity,the improvements are 5.1%,14.04%,and 7.3%,respectively,ranking first among the four types of fusion methods.It is shown that the method proposed in this paper effectively achieves polarization-based feature fusion and enhances the difference between the artificial target and the natural background.At the same time,the recognition accuracy of the fused polarized hyperspectral image for the target reaches 0.986 2,which is 21%higher than the target recognition accuracy of the single-intensity image.The experimental results show that the proposed method can effectively fuse the intensity and polarization information to improve image contrast and readability.The fused image also significantly improves target recognition accuracy,overcoming the problem of high false alarm rate of traditional spectral means for camouflage target recognition,and providing a new and effective means for new concept spectral camouflage disclosure,which has great application value. |
|
来源
|
光谱学与光谱分析
,2023,43(4):1254-1261 【核心库】
|
|
DOI
|
10.3964/j.issn.1000-0593(2023)04-1254-08
|
|
关键词
|
偏振光谱图像
;
NSCT
;
特征融合
;
伪装识别
|
|
地址
|
1.
中国科学院西安光学精密机械研究所, 陕西, 西安, 710119
2.
中国科学院大学, 北京, 100049
3.
中国科学院光谱成像技术重点实验室, 中国科学院光谱成像技术重点实验室, 陕西, 西安, 710119
|
|
语种
|
中文 |
|
文献类型
|
研究性论文 |
|
ISSN
|
1000-0593 |
|
学科
|
自动化技术、计算机技术 |
|
基金
|
国防科技创新特区项目
;
中国科学院战略性科技先导专项A
;
陕西省重点研发计划项目
|
|
文献收藏号
|
CSCD:7444120
|
参考文献 共
16
共1页
|
|
1.
Sun Qiuju.
红外,2016,37(1):5
|
CSCD被引
1
次
|
|
|
|
|
2.
徐文斌. 长波红外高光谱偏振特性的伪装目标识别方法.
光谱学与光谱分析,2019,39(1):235
|
CSCD被引
6
次
|
|
|
|
|
3.
陈振跃.
光子学报,2010,39(S1):43
|
CSCD被引
6
次
|
|
|
|
|
4.
Rahmani S.
IEEE Geoscience and Remote Sensing Letters,2010,7(4):746
|
CSCD被引
33
次
|
|
|
|
|
5.
Guo Qing.
Optik,2011,122(9):811
|
CSCD被引
8
次
|
|
|
|
|
6.
Wady S M A.
Advances in Space Research,2020,66(7):1507
|
CSCD被引
2
次
|
|
|
|
|
7.
吴一全. 基于非下采样Shearlet和WNMF的红外热波图像融合.
光子学报,2014,43(10):1010001
|
CSCD被引
1
次
|
|
|
|
|
8.
Lu H.
Computers & Mathematics with Applications,2012,64(5):996
|
CSCD被引
10
次
|
|
|
|
|
9.
沈薛晨. 基于小波-Contourlet变换的偏振图像融合算法.
红外技术,2020,42(2):182
|
CSCD被引
8
次
|
|
|
|
|
10.
Adu J.
Infrared Physics & Technology,2013,61:94
|
CSCD被引
24
次
|
|
|
|
|
11.
施文娟.
微型机与应用,2012,31(19):45
|
CSCD被引
1
次
|
|
|
|
|
12.
张莹. 基于NSCT的红外与可见光图像融合.
计算机工程与应用,2011,47(3):196
|
CSCD被引
9
次
|
|
|
|
|
13.
Zhang Q.
Infrared Physics & Technology,2016,74:11
|
CSCD被引
23
次
|
|
|
|
|
14.
Thanki R.
Multimedia Tools and Applications,2021(3):1
|
CSCD被引
1
次
|
|
|
|
|
15.
Ye Z.
Journal of Applied Remote Sensing,2017,11(3):035002
|
CSCD被引
6
次
|
|
|
|
|
16.
Dinesh Kumar M. A Comparative Study on CNN,BOVW and LBP for Classification of Histopathological Images.
2017 IEEE Symposium Series on Computational Intelligence(SSCI),2017:1
|
CSCD被引
1
次
|
|
|
|
|
了解气象卫星更多信息，请访问