Statistical shape modeling is a promising approach to map out the variability of a population. By adapting the parameters of the shape model, a new, realistic surface can be obtained. In this work, the framework for surface registration and building a statistical shape model of fruit is described.
The framework consists of two parts. First, a reference surface is registered to each fruit. Based on the correspondences that resulted from this surface registration, a statistical shape model is built. In the surface registration part, a reference surface is registered to a target surface, such that the geometric distance between those surfaces becomes minimal while maintaining correspondences. The second part of our framework consists of building a statistical shape model based on the correspondences that resulted from the surface registration (Fig. 1). The model is built by performing principal
components analysis on the corresponding points matrix of the population. In this model, the mean surface and the main shape modes are captured.
Parametrization of a surface is the task of defining a map between the surface and a simple parameter domain, such as a cylinder or sphere, so each fruit in the model can be described using basis functions, such as B-splines. This is a very compact representation and is useful in CAD and finite-element environments, so the models can be used for simulations.
Experiments resulted in better correspondences than the current state-of-the-art. This means that our shape model is a good representation of the population and adapting the shape model parameters will lead to a realistic surface.
Possible applications of a statistical shape model of fruit are predicting the final size of the fruit, search for correlations between stages of growth, estimate the volume from a single view and evaluation of the effect of the fruit shape on airflow characteristics to obtain cooling uniformity.