A UAV-based rut depth detection - A potential for evaluating soil compaction on farmlands

In recent decades, due to the increasing weight of agricultural machineries, undesirable soil compaction has become a severe factor for soil degradation. It does not only have negative ecological, but also economic effects. Successfully detecting, monitoring and predicting depths of ruts that were used by heavy vehicles can potentially reduce soil compaction. A crucial task is to generate real-world data at site-specific locations. UAV-based approaches have the advantage of providing sufficient spatial coverage and resolution to assess vital information, which can be used to directly evaluate the compaction resulting from the utilization of these heavy machineries. This information could be used for agricultural predictions, optimized routing and best time for treatments, soil regeneration purposes and many more. Therefore, the aim of this research was to spatially detect the depth of ruts, caused by heavy farm machineries on agricultural fields with consumer-grade Unmanned Aerial Vehicles (UAVs).

We therefore created a semi-automatic processing pipeline for UAV based data. The georeferenced RGB orthomosaic was used to spatially predict lanes, in the early stages of the crop cycle, by employing a Machine Learning approach. This prediction was subsequently used to extract the height information of the rut and the surrounding area from the SfM (Structure from Motion) Digital Elevation Model. As a reference method for the absolute height information, we compared this DEM (DJI Phantom 4) to the UAV - LiDAR derived DEM (RIEGL miniVUX-1UAV). For both systems, no substantial difference in the quality of the evaluated compaction depth was observed. This allows the use of low-cost UAV RGB systems to contribute to the ongoing research on soil compaction.