Measuring forest machine rut depth using inexpensive remote sensing methods : A case study in Finland

Abstract

Forest operations may result in rut formation detrimental to the forest environment. Affordable methods for monitoring rutting are therefore needed. In this study, three inexpensive remote sensing methods were tested for measuring rutting: a drone-based camera using photogrammetry (UAVPH); RGB-depth simultaneous localization and mapping with a mobile stereo camera (RGB-D SLAM); and mobile LiDAR scanning with an iPad (iPad). The measurements were performed at two forest operation sites (A and B) in Finland. Sufficiently reliable results were obtained with UAVPH and RGB-D SLAM on site A, which consisted of open area. Here, UAVPH and RGB-D SLAM produced rut depth estimates with a root-mean-square error (RMSE) of 4 to 7 cm. On site B, trees surrounding the ruts were present. Here, the accuracy of UAVPH was lower than on site A, with an RMSE of 12 and 14 cm for the two ruts respectively. On this site, RGB-D SLAM gave an RMSE as high as 43 and 108 cm due to lower computational power being available during measurement. Pearson’s correlation between the remote sensing measurements and reference values was over 0.90 for UAVPH and RGB-D SLAM on site A. On site B, correlation for UAVPH was over 0.70, but correlation for RGB-D SLAM was low. The iPad did not produce results of useful accuracy. With a clear view of the ruts being imaged and with sufficient computational power on site, the UAVPH and RGB-D SLAM methods appear promising approaches for monitoring rut depth in real forest operations, UAVPH being the superior of the two.