Methodology:

Data collection:

Aerial survey conducted using a RPAS (AgEagle eBee X) equipped with a AgEagle Duet T sensor. The Duet T sensor combined an RGB, 20 megapixel (MPx) S.O.D.A. (Sensor for Obstacle Detection and Avoidance) with a 640 x 512 Px thermal FLIR (Forward Looking Infrared) sensor for simultaneous thermal and RGB image collection. The aerial survey consisted of a single flight collecting overlapping imagery of the target area, with a minimum forward overlap of 80% and a side overlap of 60% (based on the specifications of the thermal sensor).Within the flight, five smaller discrete surveys were undertaken, with a single survey at KEP, followed by four repeat surveys of the same area at Grytviken. The KEP survey was conducted with a thermal Ground Sample Distance (GSD) of 10 cm, while the four repeat surveys at Grytviken were at four different GSD resolutions (8, 10, 12 and 14 cm).

During the aerial surveys a geodetic Trimble GNSS base station receiver collected data.

Post flight processing:

Precise point positioning (PPP) was undertaken using the online Canadian Spatial Reference System PPP tool to generate the position of the base station during the survey and which was then used to process the eBee X's onboard GNSS receiver using post-processed kinematic (PPK) processing to improve the accuracy of the camera centres captured during flight.

Photogrammetry / final outputs:

The photogrammetry and orthorectification workflows were undertaken in Pix4D (version 4.7.5.). The photogrammetry workflow used the higher resolution RGB imagery to derive the position and orientation of both the RGB and thermal sensors during each image capture (x, y, z, omega, phi and kappa (or roll, pitch, yaw)). For each discrete survey, this information was then used to derive a series of dense 3D point clouds of the surface from the RGB images and from which a DSM (digital surface model) and a triangulated mesh could be created. The triangulated meshes were then used to orthorectify both the RGB and thermal imagery, removing camera perspective and terrain distortions, resulting in a series of geometrically corrected, georeferenced orthorectified image mosaics.

Please refer to Coleman et al. for a more detailed description of the data collection and post processing methodologies.

Data quality:

PPP (precise point positioning) processing accuracy is reported in the CSRS (Canadian Spatial Reference System) PPP processing report (.pdf).

PPK (post-processed kinematic) processing accuracy (by image) is reported in the eMotion camera centre geoinfo file (.txt).