Testing ground-based robotics as remote-sensing platforms for Structure from Motion – implications for planetary science

Abstract

We assessed robotic performance by performing Structure from Motion (SfM) mapping surveys of a small gully in New South Wales, Australia, using two rovers of different sizes that each captured photographic images from a high-resolution, 12 megapixel (MP) camera and a low-resolution (<1 MP) camera. Results from these trials were compared with a simulated unmanned aerial vehicle (UAV) SfM survey of the same gully feature using the high-resolution camera on a boom pole. We conducted this in order to determine whether small, ground-based vehicles would be suitable remote-sensing platforms for collecting this data where UAV operations would be impractical or impossible. Accuracy of the resulting point clouds and the time taken to complete the surveys were used as the key metrics to assess the performance of the respective methods. The high-resolution camera surveys produced digital elevation models (DEMs) that corresponded closely to the control points surveyed using Real Time Kinematic (RTK) global positioning system (GPS) technology, although the robotic surveys took longer to complete. Additionally, camera resolution and height above ground were the major factors when determining the success of generating SfM data. We found that despite these limitations, ground-based vehicles are capable of generating point clouds accurate enough to be used to investigate small-scale geomorphology.