An image-based method for the morphological analysis of tendrils with 2d piece-wise clothoid approximation model

In this work, we present an image-based method based on 2D piece-wise clothoid for curvature approximation, that is used to analyse the morphology of natural tendrils. Starting from our previous work, here we present an advancement of the sorting skeletonization algorithm which now can handle abrupt changes in the direction of the extracted points. Furthermore, we present an automatic method to identify the minimum number of 2D piece-wise clothoid spirals needed to represent a given tendril. In our tests, we found that a range of 4–6 segments were enough to correctly represent curling shapes with high accuracy (𝑅2>0.9). The approach can be adopted for the morphological analysis of continuum growing structures, to gain new insights for designing and developing new intelligent robotic systems, such as controllable tendril-like soft robot for exploration of complex environments.