Hybrid Detectors for Subpixel Targets

Abstract

Subpixel detection is a challenging problem in hyperspectral imagery analysis. Since the target size is smaller than the size of a pixel, detection algorithms must rely solely on spectral information. A number of different algorithms have been developed over the years to accomplish this task, but most detectors have taken either a purely statistical or a physics-based approach to the problem. We present two new hybrid detectors that take advantage of these approaches by modeling the background using both physics and statistics. Results demonstrate improved performance over the well known AMSD and ACE subpixel algorithms in experiments that include multiple targets, images, and area types--especially when dealing with weak targets in complex backgrounds.

A novel hyperspectral image clustering method based on spectral unmixing

“…Therefore, the FCLS solution provides abundance estimates that meet the linear mixing model constraints, but does not allow a closed-form mathematical solution due to the non-negativity constraints. Instead, a numerical solution is proposed [9]. In fact, FCLS aims at minimizing the least squares error (LSE), which can be expressed as…”

Fuzzy Spectral and Spatial Feature Integration for Classification of Nonferrous Materials in Hyperspectral Data

“…Once the hyperspectral data is decorrelated, the feature vectors describing the spectrum are extracted and used for classification tasks. Typically, the feature vectors are constructed using the spectral information associated with each pixel in the image [29]- [31], but recent approaches suggested to merge the spectral and spatial information in order to increase the classification accuracy [32]- [34].…”

Context dependent spectral unmixing.

“…Unlike OSP and AMSD, which perform an unconstrained least squares estimate of the abundances using orthogonal projection [76], the Hybrid Subspace Detector (HSD) estimates these abundances using the fully constrained least squares algorithm [79]. This assures the satisfaction of the non-negativity and sum-to-one constraints and provides a meaning to the abundances [79].…”

“…This assures the satisfaction of the non-negativity and sum-to-one constraints and provides a meaning to the abundances [79]. HSD uses the following detection statistic:…”

“…In this section, we extend the AMSD [76], OSP [76] and HSD [79] to multiple contexts using RCDSU. A test pixel, x j , will be assigned a local detection statistic within each context i, and these local statistics will be combined using the pixel's fuzzy memberships resulting from RCDSU, u ij , to produce one confidence value.…”