Analyses of plants’ geometrical shape is of great value for many precision agriculture methodologies. Among them is the estimation of growth parameters which provide the basis for biological modeling and site-specific management. Single-image 2D-based analysis is the commonly applied approach for parameter estimation, but its accuracy is affected by imaging position, plant density (e.g., overlapping canopies), and species that share similar canopy architecture. With today’s rapid increase in computational power, stereovision modeling has become an attractive alternative for providing detailed 3D plant models. Nonetheless, the existing modeling approaches are limited in handling multiple species and growth stages, and their accuracy is affected by outdoor illumination. Moreover, they can only provide directly estimated parameters (height and leaf cover), whereas the important matter of biomass is ignored. This study proposes a novel approach for 3D plant modeling. The reconstruction stage of the model integrates local and global optimization criteria, which enables handling the challenging low textures inherent to plant scenes. In addition, it uses hue-invariant transformation for plant extraction, which has been proven robust for field illuminations. The model provides a detailed 3D reconstruction of plants’ shapes as a basis for estimating their growth parameters, including biomass. The generalized nature of its performance was proven by reconstructing the geometric shapes of different plant species at different growth stages, from young seedlings to fully developed plants. Its generalized use does not require any particular setups or adaptations, and accurate estimations of plant height (error ∼4%) and leaf cover area (error ∼4.5%) were obtained. Furthermore, a strong correlation (R2∼0.94) was found between the plant’s measured biomass and its estimated volume, which provided an accurate estimate of biomass (error ∼4%) in the validation tests. Since the proposed 3D modeling approach is inexpensive, accessible and efficiently processed, it can be implemented from agricultural vehicles for real-time applications.