Estimation of Plants’ Growth Parameters via Image‐Based Reconstruction of Their Three‐Dimensional Shape

Biomass, height, and leaf cover are indicative parameters of a plant’s growth stage and physiological condition, and their estimation establishes the basis for biological modeling and application of precision‐agriculture practices. Autonomous estimation of these parameters can be obtained via single‐image, two‐dimensional analysis, but accuracy is subject to imaging setups. Rapid increases in computational power have made stereovision models (i.e., three‐dimensional geometrical models obtained from two images) an attractive alternative by yielding detailed plant models. Nonetheless, the three‐dimensional models that have been proposed are limited in their application because of the inherent difficulty in modeling the plant’s complex shapes using only radiometric information; moreover, the important aspect of biomass estimation has been ignored. This study evaluated the potential of a three‐dimensional plant‐oriented stereovision model for biomass estimations and its additive value compared with commonly used methods. The three‐dimensional reconstruction part of the algorithm integrates local and global optimization criteria, which enables handling low‐textured plant scenes. It reconstructs the geometric shape of plants with no particular setups or adaptations. The algorithm was tested on different plants species, from young seedlings to fully developed growth stages, and accurately estimated their height (error ∼ 4%) and leaf cover area (error ∼ 4.5%). Furthermore, a strong correlation ( R 2 ∼ 0.94) was found between the plant’s estimated volume and measured biomass, providing an accurate biomass estimator in the validation tests (error ∼ 4%). Biomass estimations remained accurate at various plant densities, imaging positions, and illumination, suggesting an advantage of the three‐dimensional modeling approach over the commonly used one.