森林作業道におけるSemantic Segmentationに基づく形状情報を考慮した走路検出

抄録

Smart forestry is promoted, efficient timber transportation using forestry vehicles like forwarders is demanded. However, forest roads with steep cliffs and narrow paths pose significant safety risks, including falls and collisions. This study proposes a method to detect safe areas where vehicles can run to enhance safety in works. In proposed method, a wide-angle camera is mounted on a vehicle to recognize runnable area ahead using semantic segmentation. The proposed method was validated through experiments on actual forest roads with a forwarder. Its performance was confirmed by comparing the results with judgments by experienced forest workers.