A Comparison of Digital Vegetation Mapping and Image Orthorectification Methods Using Aerial Photography of Valley Forge National Historical Park

Show full item record

Title: A Comparison of Digital Vegetation Mapping and Image Orthorectification Methods Using Aerial Photography of Valley Forge National Historical Park
Author: Koch, Frank Henry Jr.
Advisors: Dr. Heather Cheshire, Chair
Dr. Hugh Devine, Member
Dr. Thomas Wentworth, Member
Abstract: In recent years, mapping software utilizing scanned—or "softcopy"— aerial photographs has become widely available. Using scanned photos of Valley Forge (PA) National Historical Park, I explored some of the latest tools for image processing and computer-based vegetation mapping. My primary objective was to compare different approaches for their efficiency and accuracy. In keeping with the USGS-NPS Vegetation Mapping Program protocol, I classified the park's vegetation according to The Nature Conservancy's National Vegetation Classification System (NVCS). Initially, I scanned forty-nine 1:6000 color-infrared air photos of the area at 600 dpi using an Epson desktop scanner. I orthorectified the images by two different methods. First, I did so on a single-image basis using ERDAS Imagine. In this approach, United States Geological Survey (USGS) Digital Ortho Quarter Quadrangles (DOQQ) and a 10-meter Digital Elevation Model (DEM) served as references for between seven and twelve ground control points per photo. After achieving a root mean square error (RMSE) of less than 1 meter for an image, I resampled it into an orthophoto. I then repeated the process using Imagine Orthobase. Via aerial triangulation, Orthobase generated an RMSE solution for the entire block of images, which I resampled into orthophotos using a batch process. Positional accuracies were remarkably similar for image mosaics I created from the single-image as well as the Orthobase orthophotos. For both mosaics, planimetric x-coordinate accuracy met the U.S. National Map Accuracy Standard for Class 1 maps, while planimetric y-coordinate accuracy met the Class 2 standard. However, the Orthobase method is faster—reducing process time by 50%—and requires 20% (or less) of the ground control points necessary for the single-image method. I delineated the park's vegetation to the formation level of the NVCS. Using ESRI ArcMap, I digitized polygons of homogeneous areas observed from the orthophotos. This on-screen mapping approach was largely monoscopic, though I verified some areas using a scanning stereoscope and the original hard-copy photos. The minimum mapping unit (MMU) was 0.5 acres (ac), smaller than that recommended by the USGS-NPS protocol. Based on field data, thematic accuracy for this map met the National Map Accuracy Standard of 80%. Misestimation of the hydrologic period of certain polygons resulted in some classification errors, as did confusion between evergreen and deciduous vegetation. In addition to orthophotos, Orthobase creates a stereo block viewable in ERDAS Stereo Analyst, a digital stereoscopic software package. Using Crystal Eyes™ eyewear and a high-refresh-rate monitor, a user can observe imagery full screen, three-dimensionally. Features delineated on the images are stored in ESRI shapefile format. I created a preliminary vegetation map at the alliance level of the NVCS with this software. Thematic accuracy of this map will be known when assessment is completed this summer. Notably, the classification scheme has required revision to accommodate the anthropogenically altered landscape of Valley Forge. Nevertheless, it is clear that Stereo Analyst offers advantages for vegetation and other types of mapping. Stereoscopic view and sharp zoom-in capabilities make photo interpretation straightforward. Because features are delineated directly into a GIS, Stereo Analyst cuts process time by 70% and avoids two steps that can introduce errors in conventional mapping methods (i.e., transfer to map base and digitizing). Perhaps most importantly, joint use of Orthobase and Stereo Analyst allows simultaneous orthophoto creation and GIS data collection; in contrast, the ArcMap approach requires finished orthophotos before features can be delineated. Ultimately, though, both monoscopic and stereoscopic methods have roles in vegetation mapping projects. The level of detail required for the project should determine which technique is most appropriate.
Date: 2001-04-18
Degree: MS
Discipline: Natural Resources
URI: http://www.lib.ncsu.edu/resolver/1840.16/2670


Files in this item

Files Size Format View
etd.pdf 10.07Mb PDF View/Open

This item appears in the following Collection(s)

Show full item record