基于子空间划分和耀斑抑制的赤潮高光谱高精度快速检测研究
|
摘要
高光谱数据波段多、数据量大,波段间存在信息冗余,影响了赤潮高光谱遥感检测速度。又由于海面太阳耀斑的存在,导致高光谱图像耀斑区的光谱反射远大于周边水体的离水辐亮度,严重影响了赤潮高光谱遥感检测精度。文中提出了互信息子空间划分结合最大信息熵的特征波段提取方法 ME,并与MP、CE和CP 3种方法进行对比分析,采用了3种滤波方法和6个滤波窗口进行耀斑抑制,探讨耀斑对赤潮高光谱遥感检测精度的影响。结果表明:提出的ME方法不仅具有良好的保谱性质,而且在保持原有赤潮检测精度的基础上,相较于原始全波段高光谱数据减少了近一半的检测时间;7×7滤波窗口的Median滤波方法对应的赤潮检测精度最高,达到了96.25%,相比未经滤波高光谱图像的检测精度提高了6.33%。因此,基于子空间划分的特征波段提取和耀斑抑制可实现赤潮高光谱高精度快速检测。
Hyperspectral data has many bands and large amounts of data. There is information redundancy between the bands, which affects the detection speed of red tide hyperspectral remote sensing. Due to the existence of sun glint on the sea surface, the spectral reflection in the glint area of the hyperspectral image is far greater than that of the surrounding water, which seriously affects the accuracy of the hyperspectral remote sensing detection of red tide. This paper proposes a feature band extraction method for mutual information subspace and maximum information entropy, which is compared with three methods: MP, CE and CP. Three kinds of filtering methods and six filtering windows were used to suppress the glint. The influence of glint on the hyperspectral remote sensing detection accuracy of red tide is discussed, and the results show that the ME method proposed in this paper not only has good spectral preservation properties, but also can reduce the detection time by nearly half compared with the original full-band hyperspectral data on the basis of maintaining the original red tide detection accuracy; The Median filtering method of 7 ×7 filter window has the highest detection accuracy of red tide, reaching 96.25%, which is 6.33% higher than that of unfiltered hyperspectral image. Therefore, red tide hyperspectral high-precision and rapid detection can be realized based on the subspace partition feature band extraction and glint suppression.
Hyperspectral data has many bands and large amounts of data. There is information redundancy between the bands, which affects the detection speed of red tide hyperspectral remote sensing. Due to the existence of sun glint on the sea surface, the spectral reflection in the glint area of the hyperspectral image is far greater than that of the surrounding water, which seriously affects the accuracy of the hyperspectral remote sensing detection of red tide. This paper proposes a feature band extraction method for mutual information subspace and maximum information entropy, which is compared with three methods: MP, CE and CP. Three kinds of filtering methods and six filtering windows were used to suppress the glint. The influence of glint on the hyperspectral remote sensing detection accuracy of red tide is discussed, and the results show that the ME method proposed in this paper not only has good spectral preservation properties, but also can reduce the detection time by nearly half compared with the original full-band hyperspectral data on the basis of maintaining the original red tide detection accuracy; The Median filtering method of 7 ×7 filter window has the highest detection accuracy of red tide, reaching 96.25%, which is 6.33% higher than that of unfiltered hyperspectral image. Therefore, red tide hyperspectral high-precision and rapid detection can be realized based on the subspace partition feature band extraction and glint suppression.
引文
[1]丘仲锋,崔廷伟,何宜军.基于水体光谱特性的赤潮分布信息MODIS遥感提取[J].光谱学与光谱分析,2011,31(8):2233-2237.
[2]Moradi M,Kabiri K.Red tide detection in the Strait of Hormuz(east of the Persian Gulf)using MODIS fluorescence data[J].International Journal of Remote Sensing,2012,33(4):1015-1028.
[3]Ahn Y H,Shanmugam P.Detecting the red tide algal blooms from satellite ocean color observations in optically complex Northeast-Asia Coastal waters[J].Remote Sensing of Environment,2006,103(4):419-437.
[4]Hu C M,Feng L.Modified MODIS fluorescence line height data product to improve image interpretation for red tide monitoring in the eastern Gulf of Mexico[J].Journal of Applied Remote Sensing,2017,11(1):1-14.
[5]Shanmugam,Palanisamy.A new bio-optical algorithm for the remote sensing of algal blooms in complex ocean waters[J].Journal of Geophysical Research Oceans,2011,116:1-12.
[6]Takahashi W,Kawamura H,Omura T,et al.Detecting red tides in the Eastern Seto Inland Sea with satellite ocean color imagery[J].Journal of Oceanography,2009,65:647-656.
[7]范学炜,张汉禧,孙幸文.成像高光谱数据在赤潮监测和识别中的应用研究[J].国土资源遥感,2003(1):8-12.
[8]马毅,张杰,崔廷伟.基于SVM方法的赤潮生物优势种航空高光谱识别[J].光谱学与光谱分析,2006,12(26):2302-2305.
[9]Zhang J,Wang B,Yang H M.An algorithm for red tides species’identification based on spectral characteristics and wavelet analysis[J].Applied Mechanics and Materials,2015,719-720:1063-1067.
[10]周刚,张曼祺,李佳,等.基于SVM方法的航空高光谱赤潮信息提取与分析[J].应用海洋学学报,2014,33(3):440-446.
[11]谷延峰,张晔.基于自动子空间划分的高光谱数据特征提取[J].遥感技术与应用,2003,18(6):384-387.
[12]Heden O,Lehmann J,Nǎstase E,et al.Extremal sizes of subspace partitions[J].Designs Codes&Cryptography,2012,64(3):265-274.
[13]Nǎstase E,Sissokho P.The structure of the minimum size supertail of a subspace partition[J].Designs Codes&Cryptography,2016:1-15.
[14]Li W,Zhao Y.A novel subspace partition method for fourth-order statistics based MUSIC algorithm[C]//Applied Computational Electromagnetics Society Symposium-Italy.IEEE,2017:1-2.
[15]Wu S,Yang C,Jia C,et al.ASP:Abstraction subspace partitioning for detection of atomicity violations with an empirical study[J].IEEE Transactions on Parallel&Distributed Systems,2016,27(3):724-734.
[16]Song L P,Pasion L R,Oldenburg D W.Space-time MUSIC imaging of EMI sensing data and a subspace partition study[J].Journal of Electromagnetic Waves&Applications,2017,31(15):1-21.
[17]曹彬才,朱述龙,邱振戈,等.南海区域卫星多光谱影像太阳耀斑消除方法[J].测绘科学技术学报,2017,34(2):187-192.
[18]Amankwah A.Spatial mutual information based hyperspectral band selection for classification[J].The Scientific World Journal,2015,2015:1-7.
[2]Moradi M,Kabiri K.Red tide detection in the Strait of Hormuz(east of the Persian Gulf)using MODIS fluorescence data[J].International Journal of Remote Sensing,2012,33(4):1015-1028.
[3]Ahn Y H,Shanmugam P.Detecting the red tide algal blooms from satellite ocean color observations in optically complex Northeast-Asia Coastal waters[J].Remote Sensing of Environment,2006,103(4):419-437.
[4]Hu C M,Feng L.Modified MODIS fluorescence line height data product to improve image interpretation for red tide monitoring in the eastern Gulf of Mexico[J].Journal of Applied Remote Sensing,2017,11(1):1-14.
[5]Shanmugam,Palanisamy.A new bio-optical algorithm for the remote sensing of algal blooms in complex ocean waters[J].Journal of Geophysical Research Oceans,2011,116:1-12.
[6]Takahashi W,Kawamura H,Omura T,et al.Detecting red tides in the Eastern Seto Inland Sea with satellite ocean color imagery[J].Journal of Oceanography,2009,65:647-656.
[7]范学炜,张汉禧,孙幸文.成像高光谱数据在赤潮监测和识别中的应用研究[J].国土资源遥感,2003(1):8-12.
[8]马毅,张杰,崔廷伟.基于SVM方法的赤潮生物优势种航空高光谱识别[J].光谱学与光谱分析,2006,12(26):2302-2305.
[9]Zhang J,Wang B,Yang H M.An algorithm for red tides species’identification based on spectral characteristics and wavelet analysis[J].Applied Mechanics and Materials,2015,719-720:1063-1067.
[10]周刚,张曼祺,李佳,等.基于SVM方法的航空高光谱赤潮信息提取与分析[J].应用海洋学学报,2014,33(3):440-446.
[11]谷延峰,张晔.基于自动子空间划分的高光谱数据特征提取[J].遥感技术与应用,2003,18(6):384-387.
[12]Heden O,Lehmann J,Nǎstase E,et al.Extremal sizes of subspace partitions[J].Designs Codes&Cryptography,2012,64(3):265-274.
[13]Nǎstase E,Sissokho P.The structure of the minimum size supertail of a subspace partition[J].Designs Codes&Cryptography,2016:1-15.
[14]Li W,Zhao Y.A novel subspace partition method for fourth-order statistics based MUSIC algorithm[C]//Applied Computational Electromagnetics Society Symposium-Italy.IEEE,2017:1-2.
[15]Wu S,Yang C,Jia C,et al.ASP:Abstraction subspace partitioning for detection of atomicity violations with an empirical study[J].IEEE Transactions on Parallel&Distributed Systems,2016,27(3):724-734.
[16]Song L P,Pasion L R,Oldenburg D W.Space-time MUSIC imaging of EMI sensing data and a subspace partition study[J].Journal of Electromagnetic Waves&Applications,2017,31(15):1-21.
[17]曹彬才,朱述龙,邱振戈,等.南海区域卫星多光谱影像太阳耀斑消除方法[J].测绘科学技术学报,2017,34(2):187-192.
[18]Amankwah A.Spatial mutual information based hyperspectral band selection for classification[J].The Scientific World Journal,2015,2015:1-7.