|Home||Archived November 16, 2017||(i)|
by Ellen A. Raabe and Richard P. Stumpf
OPEN-FILE REPORT 97-287
The following report presents standard methods used at the St. Petersburg Coastal and Marine Science Center for the rectification of satellite imagery and the enhancement and normalization of dn (digital number) values in all bands to permit inter-scene comparisons. Minimal pre-processing is necessary when conducting a one-time analysis of satellite imagery for land cover within a particular area. However, when a time-series of imagery is compared or when an area encompassed by many images is analyzed, accurate and compatible spatial and spectral data must be employed.
The report is divided into two chapters. Chapter 1 is written as a generic approach to image processing with explanations of how and why particular steps are conducted. Chapter 2 presents image pre-processing, step-by-step, as would be conducted with PCI, Inc. software on Landsat satellite imagery. The imagery utilized is entirely Landsat TM (Thematic Mapper) and MSS (Multi-spectral Scanner).
The first chapter includes basic image pre-processing steps in a generic format with equations for indices and calibrations. The theory and general approach may be applied to other imagery types and software. Although there will be specific references to the particulars of our own project and resources, these techniques may be creatively adapted to suit other research needs. Appendix A provides an outline of the pre-processing steps.
The second chapter covers image processing with PCI, Inc., as conducted on the Florida Big Bend coastal wetlands project. The chapter addresses software-specific parameters for each procedure, details on implementation, and includes a flow chart for PCI EASI procedures. The procedures included will permit the subsequent pre-processing of imagery in a manner suitable for additional time-series analysis.
Satellite imagery used in the detection of change along coastlines is processed in a standardized fashion to ensure temporal, spatial, and spectral compatibility between scenes. Imagery is initially selected to correlate as closely as possible with season and time-of-year coincident with high biomass and favorable atmospheric conditions as appropriate per region. Tide level at time of overpass in coastal areas can be critical in the evaluation of between-scene differences in the intertidal zone. However, due to cloud cover and image availability, tide level variations are accepted as an inherent limitation to inter-scene compatibility.
Imagery is registered to the UTM coordinate system using GPS-collected (Global Positioning System) ground control points, achieving inter-scene compatibility of < one pixel for Landsat TM. All images meet map accuracy standards of 1:25,000. Radiometric calibration and solar atmospheric correction are performed on all images to derive surface reflectance dn values in each band. The calculation of surface reflectance permits change analysis to be based on relative spectral changes or changes in vegetation biomass with less emphasis placed on land cover category variations. A vegetation index and wetness index is calculated for each image. The resulting pre-processed imagery is ready for inter-scene comparisons.
The procedures used here represent an
implementation of techniques by Dobson et al.
(1995), a document which describes approaches
recommended by the NOAA Coastal Change
Analysis Program (C-CAP).
Coastal and Marine Program > St. Petersburg Coastal and Marine Science Center > Research by Theme > Gulf of Mexico Tidal Wetlands > Image Processing Methods - OFR 97-287
|Home||Archived November 16, 2017|