Semantic Plant Segmentation Tool
A versatile and user-friendly software for accurate plant image segmentation and other foreground–background segmentation tasks using color space thresholding.
The software was developed for the computation of projected leaf area from segmented leaves or plants, but it is also applicable to other segmentation task, were a separation into foreground and background is required. The software comes with a convenient graphical user interface that gives the user direct feedback on any parametrization changes. The segmentation is based on color channel thresholding in various color spaces (HSV, L*a*b*, RGB, L*a*b*-CH, ExGR greenness). Two pre-processing options (color artifacts filter and edge-preserving blurring), two noise filters (hole filling and fragment removal) and two region-of-interest filters help to improve the segmentation results. Zoom and Pan options as well as an edge display allow for a better image handling and control. Batch processing of complete folders and various outputs are included. The output includes the computed segmentations and masked foreground and background images as well as an excel sheet output that contains the foreground pixel counts of each image.