Analysis of the residual distortion and forward motion influence on the accuracy of spatial coordinates determination based on UAV survey

Abstract

The purpose of this work is to study the operation of a non-metric digital camera Canon EOS 5D Mark III installed on a DJI S1000 octocopter, regarding the accuracy of spatial coordinates determination on images, and perform the identification and analysis of errors affecting the accuracy of stereophotogrammetry survey. During the experimental part, we conducted the stereophotogrammetric and aerial surveys of the areas including marked points. This served as a source of data for creating stereo models with their subsequent processing with the use of the Delta 2 software. The catalogs of spatial coordinates of the marked points were formed according to measurements taken by the Trimble M3 DR Total Station and from stereo models. We calculated the differences and defined root-mean-square error of determining the spatial coordinates of the points on images. Considering the specifics of the marked points placement on the studied sites, we also calculated the errors of image displacements caused by terrain. Additionally, the research studied the influence of camera`s forward motion on the accuracy of survey data of unmanned aerial vehicle (UAV). The obtained results confirm the presence of residual distortion in the optical system of the Canon EOS 5D Mark III digital camera. This leads to the need to calibrate the camera for improving the accuracy of the obtained images for their further use in mapping, monitoring geomorphological processes and phenomena, creating a Digital Elevation Model, etc. Also, the study revealed the influence of forward motion of the survey camera and image displacements caused by the height difference of the survey sites on the accuracy of created stereo models. The authors proposed a configuration and created an experimental site of marked control points on the ground for calibrating a digital non-metric camera in conditions as close as possible to the real survey conditions. Considering the analyzed literary sources, it is more effective than calibration in a laboratory.