The ability of LemnaGrid to create a topological skeleton from a shape, allows us to split an imaged plant into its larger constituents: leaf/branch and stem. If we assume that a plant is imaged from the side, then our model to segment the object is:
a stem is vertically aligned
leaves/branches are attached to the stem at an angle
leaves/branches are attached to the stem at left or right
Corn Leaf Segmenter analyses the arrangement of the skeleton and determines graphically the location of a stem and leaves/branches. Four parameters controls the result:
ignore too small edges: discards disconnected bones up to a fraction of the average leaf length
close gaps: connects bones within a certain pixel distance
leaf search radius: defines the circular area to find a leaf
initial position for stem search: the default is center-bottom of the image. Change the coordinate <X, Y> to search the stem in a region of interest.
The output of Corn Leaf Segmenter is a skeletal graph, where one or more connected bones are grouped as an image object. The objects are also labelled by their types (leaf or stem, left or right). For each object a set of morphometric parameters are determined and written to the database with Skeleton Graph writer. In addition three leaf angle information are determined from the skeleton graph and recorded in the database:
The angles are in radians and their values are converted into degrees with this formula:
angle [rad] * 180 / PI = angle [deg]
A topological skeleton of a digitised plant can be used to segment the object. Corn Leaf Segmenter applies a graphical approach to analyse the arrangement of skeletal bones and joints. It determines skeletal parts that correspond to stem and leaves of the plant. The identified leaves are used to measure three leaf angles.