OPTIMAL BAND SELECTION OF HYPERSPECTRAL DATA FOR TRANSGENIC CORN IDENTIFICATION

Resistance development by insect pests to the insecticidal proteins expressed in transgenic crops would increase reliance on broad spectrum chemical insecticides subsequently reducing environmental quality and increasing worker exposure to toxic chemicals. An important component of transgenic crop management is monitoring for insect pest resistance. This becomes critical when developing a resistance management plan for more than 25 million acres of bioengineered corn and more than 5.6 million acres of cotton currently produced in the United States (Fernandez-Cornejo, J. et al., 2006). In order to monitor compliance, the EPA has been exploring the use of airborne hyperspectral imagery. This paper reports the use of an unsupervised band selection technique for hyperspectral image analysis. This fast-search statistical approach is based on high-order moments. The features extracted possibly contain higher amount of target information. Hyperspectral data consists of bands which are very similar and these bands have close moment values. Jeffries Matusita distance, a feature similarity measure, is integrated during the band selection process so that only the most distinct bands are selected. The total number of features selected for classification is estimated using thresholding. A Mahalanobis classifier is used to evaluate the effectiveness of band selection. The investigation is focused on the comparison between the accuracies for classifications obtained when (i) all the bands were included (ii) limited, selected bands were included and (iii) randomly chosen bands were included for classification.

[1]  J. Wade Davis,et al.  Statistical Pattern Recognition , 2003, Technometrics.

[2]  Keinosuke Fukunaga,et al.  Introduction to Statistical Pattern Recognition , 1972 .

[3]  M. Paoletti,et al.  Genetic Engineering with Bacillus thuringiensis and Conventional Approaches for Insect Resistance in Crops , 2004 .

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

[5]  Nirmal Keshava,et al.  Best bands selection for detection in hyperspectral processing , 2001, 2001 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.01CH37221).

[6]  T. M. Lillesand,et al.  Remote Sensing and Image Interpretation , 1980 .

[7]  Jorge Fernandez-Cornejo,et al.  The First Decade of Genetically Engineered Crops in the United States , 2006 .

[8]  David A. Landgrebe,et al.  Signal Theory Methods in Multispectral Remote Sensing , 2003 .

[9]  M. R. Inggs,et al.  An investigation into the effects of speckle filters on classification , 1999, IEEE 1999 International Geoscience and Remote Sensing Symposium. IGARSS'99 (Cat. No.99CH36293).

[10]  R. Seidler,et al.  Application of GMOs in the U.S.: EPA research & regulatory considerations related to soil systems , 2005, Plant and Soil.

[11]  John R. Jensen,et al.  Introductory Digital Image Processing: A Remote Sensing Perspective , 1986 .

[12]  Qian Du,et al.  Band selection and its impact on target detection and classification in hyperspectral image analysis , 2003, IEEE Workshop on Advances in Techniques for Analysis of Remotely Sensed Data, 2003.

[13]  J. Hanley,et al.  The meaning and use of the area under a receiver operating characteristic (ROC) curve. , 1982, Radiology.

[14]  John A. Richards,et al.  Segmented principal components transformation for efficient hyperspectral remote-sensing image display and classification , 1999, IEEE Trans. Geosci. Remote. Sens..

[15]  Lori M. Bruce,et al.  Decision level fusion with best-bases for hyperspectral classification , 2003, IEEE Workshop on Advances in Techniques for Analysis of Remotely Sensed Data, 2003.