FIGURE 2.

Summary of the different steps allowing the detection of the adventitious roots from the images acquired from the rhizotrons. The original image (A) firstly undergoes segmentation, which allows us to obtain a binary image (B). Root skeletonization (C). For each pixel in a digital binary image, 8 neighborhoods of a pixel P are defined: 4 points in direct direction and 4 other ones in diagonal direction (Da). Different points on the skeleton can be distinguished: the endpoints (b–e), the junction points (f and g) and the normal points (h–j). Points detection (E). Identification of the adventitious root (F). Reconstruction of the adventitious root (G). Example of image of the “pruned” root (a). Histogram of the vertical radius of the right part of the root (b) thresholds (Th1 = lower threshold, Th2 = upper threshold) around a peak (c). Final image of the adventitious root without lateral information (H).