Rapid remote sensing data collection: using UAV data capture with Structure from Motion processing to generate high resolution models of seven landslide dams generated by the 2016 Kaikōura Earthquake

SKU:
SR_2018-016-pdf
$0.00
(Inc. GST)
$0.00
(Ex. GST)
Write a Review

Jones, K.E.; Lee, J.M.; Archibald, G.C.; Lawson, S.; Asher, C. 2018 Rapid remote sensing data collection: using UAV data capture with Structure from Motion processing to generate high resolution models of seven landslide dams generated by the 2016 Kaikōura Earthquake. Lower Hutt, N.Z.: GNS Science. GNS Science report 2018/16. 33 p.; doi: 10.21420/G2MH0V

Abstract: The acquisition of remotely sensed, high resolution, topographic data has been transformed considerably over the past decade by new technology and geospatial processing platforms. Originally developed within the computer vision industry, Structure from Motion (SfM) is a photogrammetric method for creating 3D models of a feature or topography from overlapping two-dimensional photographs taken from many locations and orientations. SfM algorithms search for unique matches of points between two images, determining internal camera geometry and camera position and orientation automatically, to reconstruct the photographed scene as a 3D point cloud. Estimates of motion from 2D photographs are used to rectify images into appropriate coordinates and mosaic multiple frames as well as estimate surface structure and elevation. Unmanned Aerial Vehicles (UAV) equipped with compact digital cameras can be deployed quickly and at low cost to collect high resolution images, and post processing with accurate ground control generates cm-scale orthophotos and digital surface models (DSM) suitable for detailed geomorphological mapping applications. Rapid post-event imagery capture means ephemeral features are recorded at relatively low cost for subsequent detailed geomorphological mapping applications. Following the 14 November 2016 Mw 7.8 Kaikōura earthquake landslide dams on seven rivers, Conway, Hapuku, Leader, Linton, Medway, Stanton and Towy, were flown with an UAV. Thousands of vertical and oblique overlapping photos were collected to generate high resolution, small scale digital topographic surface models of the landslide dams. Processing was undertaken using Agisoft Photoscan Professional, with manual identification of any ground control markers. Collective xyz ground control point accuracy was typically in the order of 0.2 m and as flying height dictates ground sample distance represented in the image, this varied over the high relief terrain resulting in digital surface models and orthomosaic datasets with a pixel resolution of less than 1 m. The quality, and ensuing uptake of the SfM outputs, depended on good photo overlap and regular GCP placement. (auth)