Fig. 4

Extraction and analysis of rooting traits from image data. a The positions of the manually placed markers are corrected by a local search for optimal placement of the marker. The image is first transformed by multiplication of the image with a kernel function so the regions of high pixel intensity correspond to roots that are near a marker. Local maxima are then used to nearby candidates that are placed on a root. b The path of minimum cost between a root tip and a seed is established using the Dijkstra algorithm. Because there are several seeds in the image, the network of connections between seeds and tips is pruned so that only one seed is associated to a root tip in the image. c Root skeletons, which consist of a series of paths connecting seeds and tips, are analysed to determine root traits