Support vector machines in remote sensing: A review

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2010, Volume and Issue: 66(3), P. 247 - 259

Published: Dec. 4, 2010

Language: Английский

---

Hyperspectral Unmixing Overview: Geometrical, Statistical, and Sparse Regression-Based Approaches

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Journal Year: 2012, Volume and Issue: 5(2), P. 354 - 379

Published: April 1, 2012

Imaging spectrometers measure electromagnetic energy scattered in their instantaneous field view hundreds or thousands of spectral channels with higher resolution than multispectral cameras. are therefore often referred to as hyperspectral cameras (HSCs). Higher enables material identification via spectroscopic analysis, which facilitates countless applications that require identifying materials scenarios unsuitable for classical analysis. Due low spatial HSCs, microscopic mixing, and multiple scattering, spectra measured by HSCs mixtures a scene. Thus, accurate estimation requires unmixing. Pixels assumed be few materials, called endmembers. Unmixing involves estimating all some of: the number endmembers, signatures, abundances at each pixel. is challenging, ill-posed inverse problem because model inaccuracies, observation noise, environmental conditions, endmember variability, data set size. Researchers have devised investigated many models searching robust, stable, tractable, unmixing algorithms. This paper presents an overview methods from time Keshava Mustard's tutorial present. Mixing first discussed. Signal-subspace, geometrical, statistical, sparsity-based, spatial-contextual algorithms described. Mathematical problems potential solutions Algorithm characteristics illustrated experimentally.

Language: Английский

---

Medical hyperspectral imaging: a review

Journal of Biomedical Optics, Journal Year: 2014, Volume and Issue: 19(1), P. 010901 - 010901

Published: Jan. 20, 2014

Hyperspectral imaging (HSI) is an emerging modality for medical applications, especially in disease diagnosis and image-guided surgery. HSI acquires a three-dimensional dataset called hypercube, with two spatial dimensions one spectral dimension. Spatially resolved obtained by provides diagnostic information about the tissue physiology, morphology, composition. This review paper presents overview of literature on hyperspectral technology its applications. The aim survey threefold: introduction those new to field, working reference searching specific application.

Language: Английский

---

Hyperspectral Remote Sensing Data Analysis and Future Challenges

IEEE Geoscience and Remote Sensing Magazine, Journal Year: 2013, Volume and Issue: 1(2), P. 6 - 36

Published: June 1, 2013

Hyperspectral remote sensing technology has advanced significantly in the past two decades. Current sensors onboard airborne and spaceborne platforms cover large areas of Earth surface with unprecedented spectral, spatial, temporal resolutions. These characteristics enable a myriad applications requiring fine identification materials or estimation physical parameters. Very often, these rely on sophisticated complex data analysis methods. The sources difficulties are, namely, high dimensionality size hyperspectral data, spectral mixing (linear nonlinear), degradation mechanisms associated to measurement process such as noise atmospheric effects. This paper presents tutorial/overview cross section some relevant methods algorithms, organized six main topics: fusion, unmixing, classification, target detection, parameter retrieval, fast computing. In all topics, we describe state-of-the-art, provide illustrative examples, point future challenges research directions.

Language: Английский

---

Advances in Spectral-Spatial Classification of Hyperspectral Images

Proceedings of the IEEE, Journal Year: 2012, Volume and Issue: 101(3), P. 652 - 675

Published: Sept. 10, 2012

Recent advances in spectral-spatial classification of hyperspectral images are presented this paper. Several techniques investigated for combining both spatial and spectral information. Spatial information is extracted at the object (set pixels) level rather than conventional pixel level. Mathematical morphology first used to derive morphological profile image, which includes characteristics about size, orientation, contrast structures present image. Then, neighborhood defined additional features classification. Classification performed with support vector machines (SVMs) using available postprocessing next build more homogeneous spatially consistent thematic maps. To that end, three presegmentation applied define regions regularize preliminary pixel-wise map. Finally, a multiple-classifier (MC) system produce relevant markers exploited segment image minimum spanning forest algorithm. Experimental results conducted on real different resolutions corresponding various contexts presented. They highlight importance strategies accurate validate proposed methods.

Language: Английский

---

Hyperspectral Image Classification Using Dictionary-Based Sparse Representation

IEEE Transactions on Geoscience and Remote Sensing, Journal Year: 2011, Volume and Issue: 49(10), P. 3973 - 3985

Published: May 13, 2011

A new sparsity-based algorithm for the classification of hyperspectral imagery is proposed in this paper. The relies on observation that a pixel can be sparsely represented by linear combination few training samples from structured dictionary. sparse representation an unknown expressed as vector whose nonzero entries correspond to weights selected samples. recovered solving sparsity-constrained optimization problem, and it directly determine class label test sample. Two different approaches are incorporate contextual information into recovery problem order improve performance. In first approach, explicit smoothing constraint imposed formulation forcing Laplacian reconstructed image become zero. interest has similar spectral characteristics its four nearest neighbors. second approach via joint sparsity model where pixels small neighborhood around simultaneously combinations common samples, which weighted with set coefficients each pixel. applied several real images classification. Experimental results show our outperforms classical supervised classifier support machines most cases.

Language: Английский

---

Hyperspectral Imaging: A Review on UAV-Based Sensors, Data Processing and Applications for Agriculture and Forestry

Remote Sensing, Journal Year: 2017, Volume and Issue: 9(11), P. 1110 - 1110

Published: Oct. 30, 2017

Traditional imagery—provided, for example, by RGB and/or NIR sensors—has proven to be useful in many agroforestry applications. However, it lacks the spectral range and precision profile materials organisms that only hyperspectral sensors can provide. This kind of high-resolution spectroscopy was firstly used satellites later manned aircraft, which are significantly expensive platforms extremely restrictive due availability limitations complex logistics. More recently, UAS have emerged as a very popular cost-effective remote sensing technology, composed aerial capable carrying small-sized lightweight sensors. Meanwhile, technology developments been consistently resulting smaller lighter currently integrated either scientific or commercial purposes. The sensors’ ability measuring hundreds bands raises complexity when considering sheer quantity acquired data, whose usefulness depends on both calibration corrective tasks occurring pre- post-flight stages. Further steps regarding data processing must performed towards retrieval relevant information, provides true benefits assertive interventions agricultural crops forested areas. Considering aforementioned topics goal providing global view focused hyperspectral-based supported UAV platforms, survey including sensors, inherent applications focusing agriculture forestry—wherein combination plays center role—is presented this paper. Firstly, advantages over imagery multispectral highlighted. Then, acquisition devices addressed, sensor types, modes UAV-compatible research Pre-flight operations pre-processing pointed out necessary ensure further conclusive information. With simplifying processing—by isolating common user from processes’ mathematical complexity—several available toolboxes allow direct access level-one presented. Moreover, works symbiosis between UAV-hyperspectral forestry reviewed, just before paper’s conclusions.

Language: Английский

---

Deep learning classifiers for hyperspectral imaging: A review

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2019, Volume and Issue: 158, P. 279 - 317

Published: Nov. 19, 2019

Advances in computing technology have fostered the development of new and powerful deep learning (DL) techniques, which demonstrated promising results a wide range applications. Particularly, DL methods been successfully used to classify remotely sensed data collected by Earth Observation (EO) instruments. Hyperspectral imaging (HSI) is hot topic remote sensing analysis due vast amount information comprised this kind images, allows for better characterization exploitation surface combining rich spectral spatial information. However, HSI poses major challenges supervised classification high dimensionality limited availability training samples. These issues, together with intraclass variability (and interclass similarity) –often present data– may hamper effectiveness classifiers. In order solve these limitations, several DL-based architectures recently developed, exhibiting great potential interpretation. This paper provides comprehensive review current-state-of-the-art classification, analyzing strengths weaknesses most widely classifiers literature. For each discussed method, we provide quantitative using well-known scenes, thus providing an exhaustive comparison techniques. The concludes some remarks hints about future application techniques classification. source codes are available from: https://github.com/mhaut/hyperspectral_deeplearning_review.

Language: Английский

---

Spectral–Spatial Hyperspectral Image Segmentation Using Subspace Multinomial Logistic Regression and Markov Random Fields

IEEE Transactions on Geoscience and Remote Sensing, Journal Year: 2011, Volume and Issue: 50(3), P. 809 - 823

Published: Aug. 31, 2011

This paper introduces a new supervised segmentation algorithm for remotely sensed hyperspectral image data which integrates the spectral and spatial information in Bayesian framework. A multinomial logistic regression (MLR) is first used to learn posterior probability distributions from information, using subspace projection method better characterize noise highly mixed pixels. Then, contextual included multilevel Markov-Gibbs Markov random field prior. Finally, maximum posteriori efficiently computed by min-cut-based integer optimization algorithm. The proposed approach experimentally evaluated both simulated real sets, exhibiting state-of-the-art performance when compared with recently introduced classification methods. integration of methods MLR algorithm, combined use spatial-contextual represents an innovative contribution literature. shown provide accurate characterization imagery domain.

Language: Английский

---

Advances in Hyperspectral Image Classification: Earth Monitoring with Statistical Learning Methods

IEEE Signal Processing Magazine, Journal Year: 2013, Volume and Issue: 31(1), P. 45 - 54

Published: Dec. 9, 2013

Hyperspectral images show similar statistical properties to natural grayscale or color photographic images. However, the classification of hyperspectral is more challenging because very high dimensionality pixels and small number labeled examples typically available for learning. These peculiarities lead particular signal processing problems, mainly characterized by indetermination complex manifolds. The framework learning has gained popularity in last decade. New methods have been presented account spatial homogeneity images, include user's interaction via active learning, take advantage manifold structure with semisupervised extract encode invariances, adapt classifiers image representations unseen yet scenes. This tutuorial reviews main advances remote sensing through illustrative examples.

Language: Английский

---