Quantification of Rut Detection and Height Mapping in Winter Terrains for Off-Road Mobility

Off-road autonomous vehicle navigation in winter environments requires reliable identification and quantification of potential obstacles, such as deep vehicle rutting or buried objects. The advent of consumer-grade light detection and ranging (LiDAR) sensors and unmanned aerial system (UAS) based photogrammetry present new avenues for the implementation of change detection algorithms for the purpose of obstacle identification. Few studies have provided a quantifiable statistical method for determining the input parameters of these change detection algorithms based upon user-defined confidence metrics. Previous detection methods also fail to derive the degree of assurance associated with the identification of a perceived obstacle. Here, we present an automated method for identification of snow-covered obstacles and vehicle ruts within LiDAR-derived digital elevation models based on false-alarm and detection probabilities. Detection maps and accurate height maps are generated for snow-covered objects by the algorithm to demonstrate the reliability of this method to assist with obstacle avoidance in snowy off-road conditions. The algorithm described here is a reliable and fast method for the identification and measurement of snow-covered obstacles. While this study is concerned with snow-covered terrain, the methods described here may be leveraged to monitor route deformation features as a result of vehicle traffic across a variety of terrain types.