The role of spectral resolution and classifier complexity in the analysis of hyperspectral images of forest areas.

Abstract Remote sensing hyperspectral sensors are important and powerful instruments for addressing classification problems in complex forest scenarios, as they allow one a detailed characterization of the spectral behavior of the considered information classes. However, the processing of hyperspectral data is particularly complex both from a theoretical viewpoint [e.g. problems related to the Hughes phenomenon (Hughes, 1968) and from a computational perspective. Despite many previous investigations that have been presented in the literature on feature reduction and feature extraction in hyperspectral data, only a few studies have analyzed the role of spectral resolution on the classification accuracy in different application domains. In this paper, we present an empirical study aimed at understanding the relationship among spectral resolution, classifier complexity, and classification accuracy obtained with hyperspectral sensors for the classification of forest areas. We considered two different test sets characterized by images acquired by an AISA Eagle sensor over 126 bands with a spectral resolution of 4.6 nm, and we subsequently degraded its spectral resolution to 9.2, 13.8, 18.4, 23, 27.6, 32.2 and 36.8 nm. A series of classification experiments were carried out with bands at each of the degraded spectral resolutions, and bands selected with a feature selection algorithm at the highest spectral resolution (4.6 nm). The classification experiments were carried out with three different classifiers: Support Vector Machine, Gaussian Maximum Likelihood with Leave-One-Out-Covariance estimator, and Linear Discriminant Analysis. From the experimental results, important conclusions can be made about the choice of the spectral resolution of hyperspectral sensors as applied to forest areas, also in relation to the complexity of the adopted classification methodology. The outcome of these experiments are also applicable in terms of directing the user towards a more efficient use of the current instruments (e.g. programming of the spectral channels to be acquired) and classification techniques in forest applications, as well as in the design of future hyperspectral sensors.

[1]  Christopher D. Elvidge,et al.  Comparison of relative radiometric normalization techniques , 1996 .

[2]  R. Fisher.  THE USE OF MULTIPLE MEASUREMENTS IN TAXONOMIC PROBLEMS , 1936 .

[3]  F. D. Garber,et al.  The Quality of Training Sample Estimates of the Bhattacharyya Coefficient , 1990, IEEE Trans. Pattern Anal. Mach. Intell..

[4]  Lorenzo Bruzzone,et al.  A new search algorithm for feature selection in hyperspectral remote sensing images , 2001, IEEE Trans. Geosci. Remote. Sens..

[5]  D. Sims,et al.  Relationships between leaf pigment content and spectral reflectance across a wide range of species, leaf structures and developmental stages , 2002 .

[6]  Anatoly A. Gitelson,et al.  THE SPECTRAL CONTRIBUTION OF CAROTENOIDS TO LIGHT ABSORPTION AND REFLECTANCE IN GREEN LEAVES , 2000 .

[7]  J. Peñuelas,et al.  The reflectance at the 950–970 nm region as an indicator of plant water status , 1993 .

[8]  Clement Atzberger,et al.  LAI and chlorophyll estimation for a heterogeneous grassland using hyperspectral measurements , 2008 .

[9]  Sidney Marks,et al.  Discriminant Functions When Covariance Matrices are Unequal , 1974 .

[10]  Vladimir Vapnik,et al.  Statistical learning theory , 1998 .

[11]  David G. Stork,et al.  Pattern Classification , 1973 .

[12]  D. Roberts,et al.  Hyperspectral discrimination of tropical rain forest tree species at leaf to crown scales , 2005 .

[13]  Mary E. Martin,et al.  Determining Forest Species Composition Using High Spectral Resolution Remote Sensing Data , 1998 .

[14]  S. Tarantola,et al.  Designing a spectral index to estimate vegetation water content from remote sensing data: Part 1 - Theoretical approach , 2002 .

[15]  M. Ashton,et al.  Hyperion, IKONOS, ALI, and ETM+ sensors in the study of African rainforests , 2004 .

[16]  J. Friedman.  Regularized Discriminant Analysis , 1989 .

[17]  Josef Kittler,et al.  Floating search methods in feature selection , 1994, Pattern Recognit. Lett..

[18]  E. B. Knipling.  Physical and physiological basis for the reflectance of visible and near-infrared radiation from vegetation , 1970 .

[19]  Elizabeth Pattey,et al.  Impact of nitrogen and environmental conditions on corn as detected by hyperspectral reflectance , 2002 .

[20]  David A. Landgrebe,et al.  Lowpass filter for increasing class separability , 1998, IGARSS '98. Sensing and Managing the Environment. 1998 IEEE International Geoscience and Remote Sensing. Symposium Proceedings. (Cat. No.98CH36174).

[21]  Lorenzo Bruzzone,et al.  A Novel Transductive SVM for Semisupervised Classification of Remote-Sensing Images , 2006, IEEE Transactions on Geoscience and Remote Sensing.

[22]  G. F. Hughes,et al.  On the mean accuracy of statistical pattern recognizers , 1968, IEEE Trans. Inf. Theory.

[23]  Jieping Ye,et al.  Efficient model selection for regularized linear discriminant analysis , 2006, CIKM '06.

[24]  J. Qi,et al.  A classification-based assessment of the optimal spectral and spatial resolutions for Great Lakes coastal wetland imagery , 2007 .

[25]  J. Gamon,et al.  The photochemical reflectance index: an optical indicator of photosynthetic radiation use efficiency across species, functional types, and nutrient levels , 1997, Oecologia.

[26]  A. Skidmore,et al.  Spectral discrimination of vegetation types in a coastal wetland , 2003 .

[27]  Lorenzo Bruzzone,et al.  Classification of hyperspectral remote sensing images with support vector machines , 2004, IEEE Transactions on Geoscience and Remote Sensing.

[28]  John A. Richards,et al.  Remote Sensing Digital Image Analysis , 1986 .

[29]  R. Kronmal,et al.  Discriminant functions when covariances are unequal and sample sizes are moderate , 1977 .

[30]  Lorenzo Bruzzone,et al.  An extension of the Jeffreys-Matusita distance to multiclass cases for feature selection , 1995, IEEE Trans. Geosci. Remote. Sens..

[31]  D. M. Gates,et al.  Spectral Properties of Plants , 1965 .

[32]  Lorenzo Bruzzone,et al.  Fusion of Hyperspectral and LIDAR Remote Sensing Data for Classification of Complex Forest Areas , 2008, IEEE Transactions on Geoscience and Remote Sensing.

[33]  Lorenzo Bruzzone,et al.  Semisupervised Classification of Hyperspectral Images by SVMs Optimized in the Primal , 2007, IEEE Transactions on Geoscience and Remote Sensing.

[34]  A. Gitelson,et al.  Spectral reflectance changes associated with autumn senescence of Aesculus hippocastanum L. and Acer platanoides L. leaves. Spectral features and relation to chlorophyll estimation , 1994 .

[35]  David G. Stork,et al.  Pattern Classification (2nd ed.) , 1999 .

[36]  Lorenzo Bruzzone,et al.  Kernel-based methods for hyperspectral image classification , 2005, IEEE Transactions on Geoscience and Remote Sensing.

[37]  C. Field,et al.  A narrow-waveband spectral index that tracks diurnal changes in photosynthetic efficiency , 1992 .

[38]  J. Mercer.  Functions of Positive and Negative Type, and their Connection with the Theory of Integral Equations , 1909 .

[39]  David A. Landgrebe,et al.  Covariance Matrix Estimation and Classification With Limited Training Data , 1996, IEEE Trans. Pattern Anal. Mach. Intell..

[40]  A. Gitelson,et al.  Use of a green channel in remote sensing of global vegetation from EOS- MODIS , 1996 .

[41]  Peng Gong,et al.  Land cover assessment with MODIS imagery in southern African Miombo ecosystems , 2005 .

[42]  P. Gong,et al.  Comparison of IKONOS and QuickBird images for mapping mangrove species on the Caribbean coast of Panama , 2004 .

[43]  Yong Du,et al.  Radiometric normalization, compositing, and quality control for satellite high resolution image mosaics over large areas , 2001, IEEE Trans. Geosci. Remote. Sens..