Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071804
A. Fraeman, B. Ehlmann, G. Northwood-Smith, Yang Liu, M. Wadhwa, R. Greenberger
We use VSWIR microimaging spectroscopy to survey the spectral diversity of HED meteorites at 80-μm/pixel spatial scale. Our goal in this work is both to explore the emerging capabilities of microimaging VSWIR spectroscopy and to contribute to understanding the petrologic diversity of the HED suite and the evolution of Vesta. Using a combination of manual and automated hyperspectral classification techniques, we identify four major classes of materials based on VSWIR absorptions that include pyroxene, olivine, Fe-bearing feldspars, and glass-bearing/featureless materials. Results show microimaging spectroscopy is an effective method for rapidly and non-destructively characterizing small compositional variations of meteorite samples and for locating rare phases for possible follow-up investigation. Future work will include incorporating SEM/EDS results to quantify sources of spectral variability and placing observations within a broader geologic framework of the differentiation and evolution of Vesta.
{"title":"Using VSWIR microimaging spectroscopy to explore the mineralogical diversity of HED meteorites","authors":"A. Fraeman, B. Ehlmann, G. Northwood-Smith, Yang Liu, M. Wadhwa, R. Greenberger","doi":"10.1109/WHISPERS.2016.8071804","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071804","url":null,"abstract":"We use VSWIR microimaging spectroscopy to survey the spectral diversity of HED meteorites at 80-μm/pixel spatial scale. Our goal in this work is both to explore the emerging capabilities of microimaging VSWIR spectroscopy and to contribute to understanding the petrologic diversity of the HED suite and the evolution of Vesta. Using a combination of manual and automated hyperspectral classification techniques, we identify four major classes of materials based on VSWIR absorptions that include pyroxene, olivine, Fe-bearing feldspars, and glass-bearing/featureless materials. Results show microimaging spectroscopy is an effective method for rapidly and non-destructively characterizing small compositional variations of meteorite samples and for locating rare phases for possible follow-up investigation. Future work will include incorporating SEM/EDS results to quantify sources of spectral variability and placing observations within a broader geologic framework of the differentiation and evolution of Vesta.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"83 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116313988","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071701
Junshu Wang, Guoming Zhang, Min Cao, Nan Jiang
The contradiction between high dimensional data and limited training samples is the main problem in hyperspectral remote sensing images classification. How to obtain high classification accuracy with limited labeled samples is an urgent issue. We propose a semisupervised classification algorithm SSP_EMP for hyperspectral remote sensing images based on spectral and spatial information. The spatial information is extracted by building extended morphological profiles (EMP) based on principle components of hyperspectral image. Utilize spectral and EMP from two view to enrich knowledge, and integrate the useful information of unlabeled data at the most extent to optimize the classifier. Pick high confident samples to augment training set and retrain the classifier. This process is performed iteratively. The proposed algorithm is tested on AVIRIS Indian Pines. Experimental results show significant improvements in terms of accuracy and kappa coefficient compared with the classification results based on spectral, EMP and the combination of spectral and EMP.
{"title":"Semi-supervised classification of hyperspectral image based on spectral and extended morphological profiles","authors":"Junshu Wang, Guoming Zhang, Min Cao, Nan Jiang","doi":"10.1109/WHISPERS.2016.8071701","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071701","url":null,"abstract":"The contradiction between high dimensional data and limited training samples is the main problem in hyperspectral remote sensing images classification. How to obtain high classification accuracy with limited labeled samples is an urgent issue. We propose a semisupervised classification algorithm SSP_EMP for hyperspectral remote sensing images based on spectral and spatial information. The spatial information is extracted by building extended morphological profiles (EMP) based on principle components of hyperspectral image. Utilize spectral and EMP from two view to enrich knowledge, and integrate the useful information of unlabeled data at the most extent to optimize the classifier. Pick high confident samples to augment training set and retrain the classifier. This process is performed iteratively. The proposed algorithm is tested on AVIRIS Indian Pines. Experimental results show significant improvements in terms of accuracy and kappa coefficient compared with the classification results based on spectral, EMP and the combination of spectral and EMP.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"102 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123582540","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071772
Xing Zhang, G. Wen, Bingwei Hui, Wei Dai
The aim of segmentation is to partition the image into a set of adjacent homogeneous regions. Most of existing hyperspectral imagery (HSI) segmentation approaches were designed to assign each pixel to one of the regions. However, due to the low-spatial-resolution, pixel mixing presents a challenge for HSI segmentation because a mixed spectrum does not correspond to any single well-defined material. As a result, it is difficult to determine which region the mixed pixels belong to. To address such problem, we proposed a batch-wise segmentation algorithm for HSI. First, pure pixels and mixed pixels in the HSI are separated. Then, those pure pixels are grouped into different regions. Finally, the mixed pixels are determined by its spatial neighboring pure pixels. Experimental results on a real HSI data indicate that the proposed algorithm provides more accurate segmentation maps, when compared to the traditional segmentation techniques.
{"title":"A batch-wise segmentation algorithm for hyperspectral images","authors":"Xing Zhang, G. Wen, Bingwei Hui, Wei Dai","doi":"10.1109/WHISPERS.2016.8071772","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071772","url":null,"abstract":"The aim of segmentation is to partition the image into a set of adjacent homogeneous regions. Most of existing hyperspectral imagery (HSI) segmentation approaches were designed to assign each pixel to one of the regions. However, due to the low-spatial-resolution, pixel mixing presents a challenge for HSI segmentation because a mixed spectrum does not correspond to any single well-defined material. As a result, it is difficult to determine which region the mixed pixels belong to. To address such problem, we proposed a batch-wise segmentation algorithm for HSI. First, pure pixels and mixed pixels in the HSI are separated. Then, those pure pixels are grouped into different regions. Finally, the mixed pixels are determined by its spatial neighboring pure pixels. Experimental results on a real HSI data indicate that the proposed algorithm provides more accurate segmentation maps, when compared to the traditional segmentation techniques.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124736936","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071721
Jan Koenig, L. Gueguen
In advance of releasing a WorldView-3 (WV-3) dataset with both VNIR and SWIR bands for research purposes, this study was conducted to provide a baseline comparison of land use/land cover (LULC) classification based on hyperspectral and 16-, 8-, and 4-bands of WV-3 imagery. We chose a well-researched area over the city center of Pavia, Italy. Results suggest that the addition of spectral information from WV-3's SWIR bands helps bridge the gap between precision/recall scores obtained with multispectral VNIR vs. hyperspectral VNIR imagery.
{"title":"A comparison of land use land cover classification using superspectral WorldView-3 vs hyperspectral imagery","authors":"Jan Koenig, L. Gueguen","doi":"10.1109/WHISPERS.2016.8071721","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071721","url":null,"abstract":"In advance of releasing a WorldView-3 (WV-3) dataset with both VNIR and SWIR bands for research purposes, this study was conducted to provide a baseline comparison of land use/land cover (LULC) classification based on hyperspectral and 16-, 8-, and 4-bands of WV-3 imagery. We chose a well-researched area over the city center of Pavia, Italy. Results suggest that the addition of spectral information from WV-3's SWIR bands helps bridge the gap between precision/recall scores obtained with multispectral VNIR vs. hyperspectral VNIR imagery.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124618204","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071766
Yongqiang Zhao, Chen Yi, Jingxiang Yang, J. Chan
Spectral information in hyperspectral imagery (HSI) directly acquired by sensors, commonly with surplus bands and redundant information, takes high memory and transmission costs, resulting in reduced spatial resolution and aggravated spectral mixture. Therefore, the desired high spectral resolution HSI can be obtained via spectral super-resolution after acquiring original HSI with lower spectral resolution but relatively higher spatial resolution. In this paper, we proposed a spectral super-resolution method based on spectral matrix factorization and dictionary learning. High and low spectral resolution HSIs are assumed to have the same spatial resolution and share the same spectral signatures. So abundances of low spectral resolution imagery can provide high spatial information, while its endmembers can supply accurate spectral characteristics. Then several high spectral resolution HSIs in 2-D forms are utilized to train a spectral dictionary which contains both high spatial resolution information and high spectral resolution information. Finally, the desired spectral enhancement results are achieved through the use of spatial fidelity constraint. Experiments on Sandigo dataset indicated the superiority of our proposed method.
{"title":"Spectral super-resolution based on matrix factorization and spectral dictionary","authors":"Yongqiang Zhao, Chen Yi, Jingxiang Yang, J. Chan","doi":"10.1109/WHISPERS.2016.8071766","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071766","url":null,"abstract":"Spectral information in hyperspectral imagery (HSI) directly acquired by sensors, commonly with surplus bands and redundant information, takes high memory and transmission costs, resulting in reduced spatial resolution and aggravated spectral mixture. Therefore, the desired high spectral resolution HSI can be obtained via spectral super-resolution after acquiring original HSI with lower spectral resolution but relatively higher spatial resolution. In this paper, we proposed a spectral super-resolution method based on spectral matrix factorization and dictionary learning. High and low spectral resolution HSIs are assumed to have the same spatial resolution and share the same spectral signatures. So abundances of low spectral resolution imagery can provide high spatial information, while its endmembers can supply accurate spectral characteristics. Then several high spectral resolution HSIs in 2-D forms are utilized to train a spectral dictionary which contains both high spatial resolution information and high spectral resolution information. Finally, the desired spectral enhancement results are achieved through the use of spatial fidelity constraint. Experiments on Sandigo dataset indicated the superiority of our proposed method.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121555173","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071693
Ke Wu, Q. Du
Mixed pixels in remotely sensed imagery degrade its value in practical use. Sub-pixel mapping is a promising technique to solve this problem. It can generate a fine resolution land cover map from coarse resolution fractional images by predicting spatial locations of land cover classes at sub-pixel scale. However, accuracy is often limited. When the scale factor is large, the sub-pixel distribution is complex. The traditional methods are carried out only by the fractions of land cover and the spatial dependence theory, which cannot satisfy the requirement of more accurate sub-pixel mapping. In this paper, a new observation model based on maximum a posteriori (MAP) estimation is proposed to improve the resolution of fractional images, followed by a fuzzy ARTMAP neural network to acquire a final sub-pixel mapping result. The proposed model is tested by a real remote sensed imagery, which can confirm the proposed method has better performance than the traditional algorithm, when the scale factor is large.
{"title":"Sub-pixel mapping of remotely sensed imagery based on maximum a posteriori estimation and fuzzy ARTMAP neural network","authors":"Ke Wu, Q. Du","doi":"10.1109/WHISPERS.2016.8071693","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071693","url":null,"abstract":"Mixed pixels in remotely sensed imagery degrade its value in practical use. Sub-pixel mapping is a promising technique to solve this problem. It can generate a fine resolution land cover map from coarse resolution fractional images by predicting spatial locations of land cover classes at sub-pixel scale. However, accuracy is often limited. When the scale factor is large, the sub-pixel distribution is complex. The traditional methods are carried out only by the fractions of land cover and the spatial dependence theory, which cannot satisfy the requirement of more accurate sub-pixel mapping. In this paper, a new observation model based on maximum a posteriori (MAP) estimation is proposed to improve the resolution of fractional images, followed by a fuzzy ARTMAP neural network to acquire a final sub-pixel mapping result. The proposed model is tested by a real remote sensed imagery, which can confirm the proposed method has better performance than the traditional algorithm, when the scale factor is large.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"175 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131269712","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071688
S. Livens, J. Blommaert, D. Nuyts, A. Sima, P. Baeck, B. Delauré
The COSI hyperspectral imaging system, suitable for small RPAS, is able to produce high resolution hyperspectral data products. By extensive inflight testing, we have identified the main challenges for achieving reliable high quality results. Based on these insights, we propose a refined radiometric calibration strategy. It uses a set of three reference targets, two grey and one colored target, which are to be measured inflight. We present on-ground measurements of the targets with COSI, as in flight measurements, demonstrating the merits of the approach are still ongoing.
{"title":"Radiometric calibration of the cosi hyperspectral RPAS camera","authors":"S. Livens, J. Blommaert, D. Nuyts, A. Sima, P. Baeck, B. Delauré","doi":"10.1109/WHISPERS.2016.8071688","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071688","url":null,"abstract":"The COSI hyperspectral imaging system, suitable for small RPAS, is able to produce high resolution hyperspectral data products. By extensive inflight testing, we have identified the main challenges for achieving reliable high quality results. Based on these insights, we propose a refined radiometric calibration strategy. It uses a set of three reference targets, two grey and one colored target, which are to be measured inflight. We present on-ground measurements of the targets with COSI, as in flight measurements, demonstrating the merits of the approach are still ongoing.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126801561","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071793
A. Psalta, V. Karathanassi, P. Kolokoussis
This paper aims at assessing the performance of the Simple Linear Iterative Clustering (SLIC) superpixel generating algorithm on hyperspectral images. Two modified versions of SLIC algorithm have been proposed. In the first, the HyperSLIC version, modifications were made to the basic algorithm in order to work with higher dimensions. In the second, the FD-SLIC version, a more complex distance measure, the fractional distance, already successfully used in the unmixing procedure was introduced. HyperSLIC was also applied on the abundance maps that are produced by the endmembers of the hyperspectral image. Algorithms have been applied on two images. Evaluation was based on visual inspection, NSE metric and “danger” maps. It has been shown that whole hyperspectral volume and fractional distance metric improves SLIC performance.
本文旨在评估简单线性迭代聚类(Simple Linear Iterative Clustering, SLIC)超像素生成算法在高光谱图像上的性能。提出了两个改进版本的SLIC算法。在第一个HyperSLIC版本中，为了处理更高的维度，对基本算法进行了修改。其次，介绍了FD-SLIC版本，一种更复杂的距离测量，分数距离，已经成功地应用于解混过程。hyperlic还应用于由高光谱图像的末端成员产生的丰度图。算法应用于两幅图像。评估基于目视检查、NSE度量和“危险”地图。研究表明，整体高光谱体积和分数距离度量提高了SLIC的性能。
{"title":"Modified versions of SLIC algorithm for generating superpixels in hyperspectral images","authors":"A. Psalta, V. Karathanassi, P. Kolokoussis","doi":"10.1109/WHISPERS.2016.8071793","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071793","url":null,"abstract":"This paper aims at assessing the performance of the Simple Linear Iterative Clustering (SLIC) superpixel generating algorithm on hyperspectral images. Two modified versions of SLIC algorithm have been proposed. In the first, the HyperSLIC version, modifications were made to the basic algorithm in order to work with higher dimensions. In the second, the FD-SLIC version, a more complex distance measure, the fractional distance, already successfully used in the unmixing procedure was introduced. HyperSLIC was also applied on the abundance maps that are produced by the endmembers of the hyperspectral image. Algorithms have been applied on two images. Evaluation was based on visual inspection, NSE metric and “danger” maps. It has been shown that whole hyperspectral volume and fractional distance metric improves SLIC performance.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114713249","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071707
K. Laakso, J. Peter, B. Rivard, R. Gloaguen
The most intense hydrothermally altered rocks in volcanogenic massive sulfide (VMS) deposit systems occur in the stratigraphically underlying feeder zone and rocks immediately adjacent to mineralization. This alteration zone is typically much larger than the mineralization itself, and hence the ability to detect such alteration by optical remote sensing can be invaluable for mineral exploration. Our investigation focuses on assessing the applicability of hyperspectral data to determine trends in hydrothermal alteration intensity in and around the Izok Lake VMS deposit in northern Canada. To this end, we linked hydrothermal alteration intensity information based on two indices, the Ishikawa (AI) and chlorite-carbonate-pyrite (CCPI), to hyperspectral field and laboratory data in three dimensions. Our results suggest that chlorite group minerals display variable chemical composition across the study area that broadly correlates with hydrothermal alteration intensity.
{"title":"Combined hyperspectral and lithogeochemical estimation of alteration intensities in a volcanogenic massive sulfide deposit hydrothermal system: A case study from Northern Canada","authors":"K. Laakso, J. Peter, B. Rivard, R. Gloaguen","doi":"10.1109/WHISPERS.2016.8071707","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071707","url":null,"abstract":"The most intense hydrothermally altered rocks in volcanogenic massive sulfide (VMS) deposit systems occur in the stratigraphically underlying feeder zone and rocks immediately adjacent to mineralization. This alteration zone is typically much larger than the mineralization itself, and hence the ability to detect such alteration by optical remote sensing can be invaluable for mineral exploration. Our investigation focuses on assessing the applicability of hyperspectral data to determine trends in hydrothermal alteration intensity in and around the Izok Lake VMS deposit in northern Canada. To this end, we linked hydrothermal alteration intensity information based on two indices, the Ishikawa (AI) and chlorite-carbonate-pyrite (CCPI), to hyperspectral field and laboratory data in three dimensions. Our results suggest that chlorite group minerals display variable chemical composition across the study area that broadly correlates with hydrothermal alteration intensity.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126857235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-08-01DOI: 10.1109/WHISPERS.2016.8071712
M. Islam, Derek T. Anderson, J. Ball, N. Younan
Multiple kernel learning (MKL) is an elegant tool for heterogeneous fusion. In support vector machine (SVM) based classification, MK is a homogenization transform and it provides flexibility in searching for high-quality linearly separable solutions in the reproducing kernel Hilbert space (RKHS). However, performance often depends on input and kernel diversity. Herein, we explore a new way to extract diverse features from hyperspectral imagery using different proximity measures and band grouping. The output is fed to ℓp-norm MKL for feature-level fusion, where larger p's are preferred for diverse vs sparse solutions. Preliminary results on benchmark data indicates that ℓp-norm MKSVM of diverse features and kernels leads to noticeable performance gain.
{"title":"Fusion of diverse features and kernels using LP-norm based multiple kernel learning in hyperspectral image processing","authors":"M. Islam, Derek T. Anderson, J. Ball, N. Younan","doi":"10.1109/WHISPERS.2016.8071712","DOIUrl":"https://doi.org/10.1109/WHISPERS.2016.8071712","url":null,"abstract":"Multiple kernel learning (MKL) is an elegant tool for heterogeneous fusion. In support vector machine (SVM) based classification, MK is a homogenization transform and it provides flexibility in searching for high-quality linearly separable solutions in the reproducing kernel Hilbert space (RKHS). However, performance often depends on input and kernel diversity. Herein, we explore a new way to extract diverse features from hyperspectral imagery using different proximity measures and band grouping. The output is fed to ℓp-norm MKL for feature-level fusion, where larger p's are preferred for diverse vs sparse solutions. Preliminary results on benchmark data indicates that ℓp-norm MKSVM of diverse features and kernels leads to noticeable performance gain.","PeriodicalId":369281,"journal":{"name":"2016 8th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117260101","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：