Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.


Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:   No Citation available.
Authors:   Scott D Noble, Ralph B Brown
Keywords:   segmentation, hyperspectral imaging, edge detection, band selection

Leaf shape is an important feature used in plant identification. The first step in automating leaf shape analysis in complex, real-world scenes is to segment individual leaves. Leaves are often occluded and overlapping, and the similar colour and texture characteristics of the leaves complicate the task of separating them. A segmentation approach based on simple edge detectors operating on narrow-waveband images from an imaging spectrophotometer was developed and tested. Band selection was done by comparing separability of leaf-leaf, vein, and leaf-overlap edge regions of multi-plant, images for four species in 115 spectral bands between 400 and 1000 nm. Testing resulted in a mean segmentation percentage of 63% using a Sobel edge detector operating on an image in a 5 nm waveband centred at 719 nm. This result compared favourably with existing research given the relative simplicity of the segmentation algorithm and complexity of the test images used.

(Download PDF)    (Export to EndNotes)