Positional accuracy in a natural resource database: comparison of a single-photo resection versus affine registration

Positional and area accuracies were calculated for digitized data taken from 1:20,000 scale aerial photographs and United States Geological Survey (USGS) 1:24,000 scale topographic maps. Positional accuracy was determined as the Euclidian distance between the digitized coordinate and the reference ground coordinate collected with global positioning systems (GPS). Area accuracy was the acreage difference between the digitized area and GPS calculated area. Three methods were employed to collect the digitized data: manual digitizing from topographic maps and aerial photographs followed by an affine transformation, and manual digitizing from aerial photographs while applying a single-photo space resection. Two study sites, one in low terrain relief and one in high terrain relief, were used to examine the effects of terrain on positional accuracies.

The single-photo space resection technique provided the most accurate positional data on both study sites. The single-photo space resection produced mean positional accuracies of 5.0 to 6.0 meters. In comparison, the uncorrected digitized photo data produced mean positional accuracies of 7.0 to 26.0 meters. The effects of terrain displacement were evident in these data sets, as the mean positional accuracy at the low-relief study site was 18.96 meters less than the corresponding accuracy at the high-relief study site. The uncorrected digitized photo data set from the high-relief study site provided the highest mean positional accuracy, 25.86 meters. The topographic map digitized data from both study sites provided mean positional accuracies below 12.0 meters, but failed to meet National Map Accuracy Standards (NMAS) for 1:20,000 scale or smaller maps.

The average area accuracy from both study sites proved to be not significantly different, regardless of the digitizing technique or terrain conditions. The average area accuracy between the two study sites differed by at most 0.05 hectares. Average percent area errors ranged from 9.96% to 11.74% on the low-relief study site and from 11.84% to 12.65% on the high-relief study site.