[an error occurred while processing this directive][an error occurred while processing this directive]
Achieving Ortho-accuracy in a GIS
How accurate does a GIS need to be? How about "ortho-accurate"? The need to solve increasingly complex geospatial problems requires more information and a greater degree of accuracy found in high-resolution data such as aerial photography. New products such as IMAGINE OrthoBASE and Stereo Analyst are affordable solutions that are designed to take advantage of high-resolution data and streamline the process of populating a GIS with accurate and current data derived from geographic imagery.
A Case in Point
When GeoGraphix Inc. converted the Rankin County, Mississippi, tax-mapping project to IMAGINE OrthoBASE (see "Updating Tax Maps in Mississippi"), they soon realized that they were going to deliver a far superior product (orthorectified imagery) than the County expected (rectified imagery).
The crucial element for success was the "on-the-fly" aerial triangulation that allowed GeoGraphix to resolve low residuals across large areas, accept the most accurate base data, and exclude the inaccurate data. The readout on the computer screen provided residual errors of all the points resolved in a block. The control points and block data were kept independent of the imagery, allowing each township range block (consisting of36 square miles) to be mathematically resolved at once.
The residual errors of each point, resulting from the least-squares bundle adjustment, were viewed both graphically and in tabular form. Geographix would then either accept or reject the aerial triangulation results, modify the input, and possibly run the block again. The graphical depiction of the residuals provided a clear picture of the points with the highest residuals (indicating accuracy) and pinpointed errors.
The points indicating errors were further analyzed and observed and eventually removed from the overall solution. The final residuals were kept below a predetermined root mean square error (RMSE), meaning that the points used for anchor points were highly accurate relative to each other and the geospatial data provided by the digital elevation model (DEM) surface.
This is in contrast to rubber-sheeting, which usually requires approximately nine points per image to be selected from features that are visually identified from a basemap and used to anchor the aerial photos. The image is then "best fit" or warped to correspond to those anchor points. Determining how well the points fit is an educated guess based on the metadata and quite often results in rectified aerial images being matched to inaccurate data.
By working in large blocks, GeoGraphix was able to verify that the selected anchor points were the most accurate controls and pinpoint anomalies in the existing geospatial data based on mathematics rather than opinion. This saved the company countless hours of manual evaluation and resulted in a seamless mosaic of highly accurate information that Rankin County's ArcInfo GIS users can incorporate into their daily work.
For more information about integrating digital photogrammetry into your GIS application, please contact ERDAS (toll-free tel.: 1-877-GO ERDAS [1-877-463-7327]) or visit ERDAS on the Web at www.erdas.com.