Navigable Areas Segmentation Method for Unmanned Surface Vehicles in Paddy Fields

An Unmanned Surface Vehicle (USV) capable of automatic navigation in paddy fields could enhance paddy rice cultivation efficiency. However, flexible navigation is challenging with rule-based methods. This study introduce a deep neural network-based navigable area binary segmentation method, utilizing images from a front camera mounted on a USV. Evaluation with a newly created dataset confirmed that this method could flexibly segment navigable areas, even in rice fields where rule-based methods struggle.