Skip to main content
ARS Home » Pacific West Area » Reno, Nevada » Great Basin Rangelands Research » Research » Publications at this Location » Publication #302585
Title: Evaluation of structure from motion for soil microtopography measurement

Author
item Weltz, Mark
item NOUWAKPO, SAVJRO - University Of Nevada
item JAMES, MIKE - Lancaster University
item Huang, Chi Hua
item CHAGAS, ISIS - Federal University - Brazil

Submitted to: The Photogrammetric Record
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/11/2014
Publication Date: 9/18/2014
Publication URL: http://www.altmetric.com/details.php?domain=onlinelibrary.wiley.com&doi=10.1111/phor.12072
Citation: Weltz, M.A., Nouwakpo, S.K., James, M.R., Huang, C., Chagas, I. 2014. Evaluation of structure from motion for soil microtopography measurement. The Photogrammetric Record. 29(147):297-316.

Interpretive Summary: Recent developments in low cost structure from motion technologies offer new opportunities for geoscientists and natural resource managers to acquire high resolution soil microtopography data at a fraction of the cost of conventional techniques. This data is essential to accurately measure and predict surface soil erosion. However, these new methodologies often lack easily accessible error metrics and are hence difficult to evaluate. In this paper, a framework was developed to evaluate a structure for motion approach for soil microtopography measurement through assessment of uncertainty sources and quantification of their potential impact on overall 3D reconstruction of the soil surface. Standard deviation of camera intrinsic parameters estimated from structure for motion self-calibration from five different soil surface models were several orders of magnitude larger than the precision expected from pattern-based camera calibration. Sensitivity analysis identified camera principal point (image projection centre) as the dominant source of calibration-induced uncertainty. Overall surface elevation values estimated from both technologies were similar in magnitude with root mean square of elevation difference of 0•2 mm. This technology when properly calibrated provides new techniques to quantify both soil erosion and deposition across the landscape in a fast and economical manner using readily available digital camera and computer software.

Technical Abstract: Recent developments in low cost structure from motion (SFM) technologies offer new opportunities for geoscientists to acquire high resolution soil microtopography data at a fraction of the cost of conventional techniques. However, these new methodologies often lack easily accessible error metrics and are hence difficult to evaluate. In this paper, a framework was developed to evaluate a SFM approach for soil microtopography measurement through assessment of uncertainty sources and quantification of their potential impact on overall 3D reconstruction. Standard deviation of camera intrinsic parameters estimated from SFM self-calibration within five different soil surface models were several orders of magnitude larger than precisions expected from pattern-based camera calibration. Sensitivity analysis identified camera principal point (image projection centre) as the dominant source of calibration-induced uncertainty. Overall surface elevation values estimated from both technologies were similar in magnitude with RMS of elevation difference of 0•2 mm. Nevertheless, the presence of deformation in either SFM or traditional photogrammetry point clouds highlights the importance of quality assurance safeguards (such as a judicious choice of control points) in SFM workflows for soil microtopography applications.