Multiclassifiers and decision fusion in the wavelet domain for exploitation of hyperspectral data

Abstract

Abstract-In this paper, the discrete wavelet transform (DWT) is employed as a preprocessing stage for a multiclassifier and decision fusion system for feature extraction and dimensionality reduction of hyperspectral data. As a result, both global and local spectral features can be exploited. Feature grouping is conducted according to wavelet decomposition levels, or scales. Each DWT decomposition level's detail coefficients are classified independently, creating a multiclassifer system. The resulting classifications are then fused using a simple majority voting scheme. The proposed target recognition system was applied to hyperspectral data for an agrictultural applications, namely detecting the presence of the often devastating disease known as soybean rust in soybean crops. The proposed approach was compared to well-known hyperspectral dimensionality reduction methods, such as stepwise linear discriminant anlaysis (LDA). When using the DWT multiclassifier system, the overall classification accuracies ranged from the high 80's to the mid 90's. When using the stepwise LDA technique the overall classification accuracies ranged from the mid 60s to the mid 90's. .

Rapid detection of agricultural food crop contamination via hyperspectral remote sensing

“…In previous work [5,6,7,8], the authors have investigated a variety of new methods for hyperspectral dimensionality reduction and classification, including spectral band grouping, wavelet coefficient feature extraction and selection, and multi-classifier and decision fusion (MCDF) techniques. The authors have found that combinations of discrete wavelet transforms (DWT) with MCDF schemes are quite powerful in exploiting hyperspectral data for classifying subtly different vegetative classes.…”

“…Thus, the DWT-MCDF approach is used in this study for detecting/characterizing chemical contaminations of corn and wheat. Details of the DWT-MCDF approach can be found in [5,6,7,8]. In this study, we used Haar mother wavelet, maximum level of wavelet decomposition, maximum likelihood classifiers, and a simple majority vote for the decision fusion.…”

Empirical Mode Decomposition based decision fusion for hyperspectral image classification

“…Son olarak, karar tümleútirme iúlemi ile sonuçlar birleútirilmektedir. [6]'da ise ayrÕk dalgacÕk dönüúümü kullanarak çoklu sÕnÕflandÕrÕcÕlar elde edilmektedir. AyrÕk dalgacÕk dönüúümü sonucu farklÕ seviyelerde oluúan alçak frekans bileúenleri birbirinden ba÷ÕmsÕz olarak DVM ile sÕnÕflandÕrÕlmaktadÕr.…”

Sensitivity of hyperspectral classification algorithms to training sample size

“…The class labels, or decisions, resulting from the bank of Recently, the authors of this paper investigated the usefulness of multi-classifiers for hyperspectral image exploitation [5], [6]. We proposed two types of methods to alleviate the small sample size problem -one based on employing a Multi-Classifier Decision Fusion (MCDF) in the raw reflectance domain , and the other employed the MCDF framework in the Discrete Wavelet Transform domain (DWT-MCDF).…”

“…LOA based preprocessing is carried out at the subspace level for further dimensionality reduction and feature optimization. Maximum-likelihood classifiers are employed for 4000 This section provides a brief explanation of MCDF and DWT-MCDF; details can be found in [5], [6]. Both MCDF and DWT-MCDF use a combination of multiple classifiers and decision fusion.…”