Agriculture Surface Edge Explorer (AgriSEE): Active reconstruction of occluded tomatoes in greenhouses

Intensive greenhouse agriculture has become one of the pillars of society's population growth. However, increasing overpopulation in both the human and animal world requires existing agriculture to become more efficient and sustainable. Automation in general, and robotics in particular, play a fundamental role in addressing these challenges, particularly for `dull', `dirty' and `dangerous' tasks. This work focuses on applying a Next Best View (NBV) algorithm, SEE+, to obtain a 3D point cloud model using a Universal Robot UR10 robot arm. The work describes how to use a camera with LiDAR technology to generate a 3D model for fruits, which will later be used for the automated harvesting inside greenhouses. The NBV algorithm classifies the point cloud to determine the next best view and can efficiently obtain a scan of the fruit for subsequent harvesting. The results show how it has been possible to determine with precision the location of a tomato, something difficult to do using traditional techniques.