3D Data Collection for Individual Plant Farming

Automated phenotyping allows for more precise tracking of the health and life cycle of each plant in a farm or garden. This enables agriculture and horticulture with less water consumption, decreased reliance on pesticides, and increased crop yield. State-of-the-art techniques for plant phenotyping include imaging the plants with color and near-infrared cameras. Because 2D data is collected, parts of the plants are often occluded. The goal of this project is to estimate the various perspectives from which 2D data should be captured in order to generate 3D models of the plants with minimal occlusions. This has been explored in photogrammetry literature, but here we develop a procedure for use with a robotic arm with constrained workspace.