Image Segmentation of Hyperspectral Imagery
Abstract
Hyperspectral imagery (HSI), a passive technique creating a large collection of images of fine resolution across the infrared spectrum is currently being considered for U.S. Army tactical applications. An important tactical application of infrared (IR) hyperspectral imagery is the detection of low-contrast targets, including those targets that may employ camouflage, concealment, and deception (CCD) techniques 1, 2. Spectral reflectivity characteristics were used for efficient segmentation between different materials such as painted metal, vegetation, and soil for visible to near IR bands in the range of 0.46-1.0 um as shown previously by Kwon et al. 3. We are currently investigating the HSI region spanning the wavelengths from 7.5 to 13.7 %m. The energy in this range of wavelengths is almost entirely emitted rather than reflected; therefore, the gray level of a pixel is a function ofthe temperature and emissivity of the object. This is beneficial because light level and reflection will not need to be considered in the segmentation. We will present results of segmentation analysis on the long-wave infrared (LWIR) hyperspectrum using a simple distance metric applied to full-band and sub-band HSI data sets, neural network-based classification approaches, and principal component analysis (PCA) applied to relative temperature profiles derived from the Spatially Enhanced Broadband Array Spectrograph System (SEBASS) database. A stepwise segmentation will be demonstrated using a back-propagation neural network, which outlines some of the difficulties in the multi-class case. Overall, these results should give an early indication of the added capability hyperspectral imagery and algorithms will bring to bear on the target acquisition problem.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 2003
- Accession Number
- ADA416758
Entities
People
- Mark Wellman
- Nassar Nasrabadi
Organizations
- United States Army Research Laboratory