Sensitivity of hyperspectral classification algorithms to training sample size

Abstract

Algorithms that exploit hyperspectral imagery often encounter problems related to the high dimensionality of thẽ at~, particularly when the amount of training data is limited. Recently, two algorithms were proposed to alleviate the small sample size problem -one is 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). This paper investigates the sensitivity of conventional single classifier based classification approaches, as well as MCDF and DWT-MCDF to varia~ions .in the amount of data employed for training the classification system. The hyperspectral data in this xperiment was obtained using an airborne hyperspectral Imager used by SpecTIRTM. The results of the experimental analysis show that for the given application, the MCDF and DWT-MCDF algorithms are significantly less sensitive than the conventional algorithms to limited training data. PCA consistently results in overall accuracies of about 35%. LDA accuracies are very high, about 75%, when there is an abundance of training data -about lOX (i.e. number of traini?g samples is 10 times the number of spectral bands); remams above 60% for training data abundances of 2X and higher; but dramatically decreases to ---20% for abundances of IX. MCDF results in accuracies ranging between 65% and 75% for training data abundance of 3X and higher, but the accuracies drop to ---60% for 2X and ---55% for IX. DWT-MCDF results in high accuracies with the least sensitivity to training data abundance. Its accuracies range between ---60-65% for abundances of IX to lOX.

Non-Uniform Random Feature Selection and Kernel Density Scoring With SVM Based Ensemble Classification for Hyperspectral Image Analysis

“…The dataset has seven classes-each representing chemical stress on a corn-crop [35]. The corn crop, grown under controlled conditions was induced with varying degrees of chemical stress.…”

“…The classification is performed using SVM with Gaussian RBF kernel for all experiments [35]. Model selection for the SVM is performed by using cross validation and grid search.…”

Data dependent adaptation for improved classification of hyperspectral imagery

“…Because all seven classes in this dataset represent the same species under varying degrees of stress, it makes it a very challenging classification problem. The reader is referred to [9] for a detailed description of this dataset. Figure 3 depicts the benefits of the proposed confusion driven adaptation when using a single classifier system (SLDA based dimensionality reduction, followed by maximum-likelihood classification -SLDA-ML), and a recently proposed multi-classifier system (MCDF).…”

The Deep Belief and Self-Organizing Neural Network as a Semi-Supervised Classification Method for Hyperspectral Data

“…Spectral imaging, especially hyperspectral imaging, has advantages in being able to distinguish features (its spectral and spatial distinguishing ability), data size, and wavebands. However, with a smaller sample size for data analysis [6,7], the curse of dimensionality, and, concurrently, linear classification faults, this technology may well show its deficiencies [8,9]. At present, there are several more mature classification methods that can be applied to hyperspectral data, including support vector machines (SVMs) and spectral angle mapping (SAM) [10][11][12][13][14][15].…”