Pabón-Ramírez, Wilma N.

Profile Picture

Filters

20081
Reset filters

Settings

Citations

Publication Search Results

Now showing 1 - 1 of 1
  • Thumbnail Image
    Publication
    Soft classification of hyperspectral imagery based on linear mixing model and supervised fuzzy logic algorithms
    (2008) Pabón-Ramírez, Wilma N.; Vélez-Reyes, Miguel; College of Engineering; Hunt, Shawn D.; Manian, Vidya; Department of Electrical and Computer Engineering; Gilbes, Fernando
    Hyperspectral Imagery (HSI) is an important technology used in remote sensing and plays an important role in environmental remote sensing because it provides valuable spectral information of the objects in the scene. Using the measured spectral signatures, it is possible to discriminate between materials in the scene for object detection, recognition or identification. Hyperspectral technologies are of great value for environmental applications where it is possible to take advantage of spectral, spatial, and radiometric resolutions. A problem for current and proposed spaceborne hyperspectral platforms is their low spatial resolution which ranges from 20 to 30m. The key problem with low spatial resolution is mixed pixels where the measured spectral signature is a combination of the contributions of the spectral signatures of the materials in the field of view in the sensor. In such cases, the high spectral resolution can be used to extract information about objects at the subpixel level by their contribution to the measured spectral signature. A common technique in HSI analysis is hard classification where each pixel is assigned to one and only one specific class. In this research work, we investigated soft classification algorithms which can consider the mixed pixel problem for image classification. Soft classifiers assign multiple classes to a single pixel using membership functions which weight the membership of the pixel into the available classes. As a result, soft classification could be used to develop models and thematic maps that are more appropriate for low resolution remote sensing imagery. This thesis presents a comparative study of soft classification algorithms based on Linear Mixing Model and supervised Fuzzy Logic classification systems as an alternative for hard classification of low spatial resolution HSI. As part of the research, we developed a Spectral Soft Classification Tool (SSCT), which should be valuable resource for image analysts because it provides soft classification outputs, visualization tools, and accuracy assessment to analyze multi/hyperspectral imagery. Remotely sensed data from HYPERION and ETM+ (LANDSAT 7) collected over Puerto Rico were used in this study.