Fig. 8.

Comparison of the original (a, h) and segmented images (b–g, i–n) on two different illumination days using the three illumination-scenario specific models: general model for all light conditions (M_ALC) and models for high light-contrast (M_HLC) and low light-contrast (M_LLC). For each scenario, the support vector machines (SVM) and decision tree segmentation model (DTSM, see Guo et al. [22]) were used for training. See the details of the segmentation accuracy in the Additional file 2, the image 2014_11_07_FWW0070706