Pietola, L. and A. J. Smucker. 2006. Elimination of non-root residue by image analysis of very fine roots. Computers and Electronics in Agriculture 53:92-97.
Hydropneumatic root separation from field soil collects organic non-root residues. This study compares manual cleaning to electronic cleaning with image analysis of carrot roots from field experiments. Washed and stained very fine roots of carrots (diameter 0.15 mm) were video-recorded by using a high-resolution robotic camera system. Fibrous root lengths and widths of roots from samples grown in fine sand (n = 160) and organic soils (n = 150) were determined by automated image processing of video-recorded images. During video-recording, some non-root residue remained with the extracted fine roots, especially in samples from organic soils, where residue materials comprised partially decomposed straw and peat. In computerized image analysis, materials with length-to-width ratios of =3:1 were deemed non-root debris and discarded electronically. The 3:1 length-to-width ratio was chosen for distinguishing long thin root images from short organic residues. Results show that more non-root residue was recovered by image analysis than was recorded by dry weight estimation after manual cleaning. Length, surface area and volume averaged 15%, 18% and 24% of non-root residue in the manually fully cleaned samples (residue estimation 0%, i.e. no visible debris recorded visually) of fine sand soils (n = 80). Length, surface area, and volume averaged 20%, 25% and 33% (r = 0.55, p < 0.0001), respectively, of non-root residue in the organic soil samples with residues <10% (g residues/g roots and residues) (n = 71). Both root and residue densities were higher at soil depths of 0 – 25 cm than in deeper subsoils, where the percentages of discarded residue were highest. Electronic non-root separation effectively eliminated short and partially decomposed organic material that was hardly visible or otherwise difficult and time-consuming to remove manually.
Associated Treatment Areas:
MethodsSign in to download PDF back to index