Abstract
This paper examines the influence of raw image preprocessing and other selected processes on the accuracy of close-range photogrammetric measurement. The examined processes and features includes: raw image preprocessing, sensor unflatness, distance-dependent lens distortion, extending the input observations (image measurements) by incorporating all RGB colour channels, ellipse centre eccentricity and target detecting. The examination of each effect is carried out experimentally by performing the validation procedure proposed in the German VDI guideline 2634/1. The validation procedure is based on performing standard photogrammetric measurements of high-accurate calibrated measuring lines (multi-scale bars) with known lengths (typical uncertainty = 5 μm at 2 sigma). The comparison of the measured lengths with the known values gives the maximum length measurement error LME, which characterize the accuracy of the validated photogrammetric system. For higher reliability the VDI test field was photographed ten times independently with the same configuration and camera settings. The images were acquired with the metric ALPA 12WA camera. The tests are performed on all ten measurements which gives the possibility to measure the repeatability of the estimated parameters as well. The influences are examined by comparing the quality characteristics of the reference and tested settings.
Highlights
Photogrammetry is a measurement method based on a large number of hardware and software components, and on the user skills
This paper examines the influence of the raw image preprocessing and other selected processes on the accuracy of close-range photogrammetric measurements
The acceptance test is based on photogrammetric measurements of calibrated lengths (300– 2200 mm), realized by 7 multi-scale bars with circular retro
Summary
Photogrammetry is a measurement method based on a large number of hardware and software components, and on the user skills. This paper examines the influence of the raw image preprocessing and other selected processes on the accuracy of close-range photogrammetric measurements. All these issues have been already described and treated in the past by various authors, but at that time, the performance of digital sensors, computers and reference gauges was often limited in comparison with today’s possibilities. Our approach to testing various influences is unique in including the factor of repeatability of the photogrammetric system performance. The raw image preprocessing consists of the following steps (this is just a selection of the main components of the workflow): dark-frame subtraction (this includes fixed pattern noise, thermal noise, hardware-specific bias etc.), linearization, vignetting correction, white colour balancing, demosaicing, colour space conversion, brightness and gamma correction, noise filtering, bit depth reduction in export, compression etc. Influence of the various ellipse target detectors is studied in Trinder et al (1995) or Quellet et al (2009)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have