Single digital-photo correction for a GIS application and error analysis
Single digital-photo correction using the collinearity condition equations for a GIS application was investigated by control and test data from digital photos and USGS 1:24,000 topographic maps and by USGS 7.5-minute DEMs of two study areas, Prentiss, North Carolina and Blacksburg, Virginia. The collinearity equations were used to remove geometric errors caused by image tilt and relief, displacement. In addition to the study of geometric error in uncorrected and corrected digital photos, the impact of error in ground control points and error in DEM data on the accuracy of rectification was analyzed in terms of errors in point position, line length, and polygon area. Computer programs for single digital-photo correction using the collinearity equations and for impact analysis of errors in ground control and DEM data were developed.
Point errors in uncorrected photos were tested by the affine transformation from UTM to photo coordinates by control points. Estimated error for the Prentiss photo was 14.599 meters and that for the Blacksburg photo was 11.252 meters. Due the poor distribution of control and test points, estimated errors in uncorrected photos were underestimated. Estimated error for the corrected photo of the Prentiss area was 11.397 meters and that for the Blacksburg area was 7.071 meters. Geometric errors in corrected digital photos were clearly not significantly greater than 12.192 meters. Error removed for the Blacksburg photo was significantly greater than zero and that for the Prentiss photo was not significantly greater than zero. Estimated removed errors were also underestimated.
Errors in point position, line length, and polygon area of corrected digital photos were positively related to error in ground control points for photo orientation and error in DEM data for registration. The accuracy of ground control points has greater impact on point position and polygon area, and the accuracy of DEM data has greater impact on line length. The point position error, line length change, and polygon area change caused by random error within ±15 meters in ground control points were within 4 meters, 0.12% and 0.3% and those caused by random error within ±10 meters in DEM data of test features were within 2 meters, 0.11% and 0.0065%. It is both practical and accurate to use ground control points obtained from USGS 7.5-minute topographic maps and USGS 7.5-minute DEMs for single digital-photo correction by the collinearity equations for most GIS applications with spatial data layers obtained from USGS 1:24,000 topographic maps or smaller-scale maps for areas such as those studied here.