Premium
Evaluation of the controls affecting the quality of spatial data derived from historical aerial photographs
Author(s) -
Walstra Jan,
Chandler Jim H.,
Dixon Neil,
Wackrow Rene
Publication year - 2010
Publication title -
earth surface processes and landforms
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.294
H-Index - 127
eISSN - 1096-9837
pISSN - 0197-9337
DOI - 10.1002/esp.2111
Subject(s) - weighting , photogrammetry , a priori and a posteriori , computer science , data mining , quality (philosophy) , workflow , data quality , statistics , artificial intelligence , mathematics , metric (unit) , engineering , medicine , philosophy , operations management , epistemology , database , radiology
This paper is concerned with the fundamental controls affecting the quality of data derived from historical aerial photographs typically used in geomorphological studies. A short review is provided of error sources introduced into the photogrammetric workflow. Data‐sets from two case‐studies provided a variety of source data and hence a good opportunity to evaluate the influence of the quality of archival material on the accuracy of coordinated points. Based on the statistical weights assigned to the measurements, precision of the data was estimated a priori, while residuals of independent checkpoints provided an a posteriori measure of data accuracy. Systematic discrepancies between the two values indicated that the routinely used stochastic model was incorrect and overoptimistic. Optimized weighting factors appeared significantly larger than previously used (and accepted) values. A test of repeat measurements explained the large uncertainties associated with the use of natural objects for ground control. This showed that the random errors not only appeared to be much larger than values accepted for appropriately controlled and targeted photogrammetric networks, but also small undetected gross errors were induced through the ‘misidentification’ of points. It is suggested that the effects of such ‘misidentifications’ should be reflected in the stochastic model through selection of more realistic weighting factors of both image and ground measurements. Using the optimized weighting factors, the accuracy of derived data can now be more truly estimated, allowing the suitability of the imagery to be judged before purchase and processing. Copyright © 2010 John Wiley & Sons, Ltd.