ROOT QUANTIFICATION BY EDGE DISCRIMINATION USING A DESK-TOP SCANNER

Root characterization methods are constrained by one or more of the following factors: cost, limited resolution, or time of analysis. Most length-determination methods are based on the line-intersect principle. An alternate method using a flatbed scanner and a microcomputer was developed. Algorithms were developed to estimate root area, length, and diameter (mean root width). Two-dimensional area is estimated by summing the number of pixels representing a single or group of gray scales. Length and width are calculated as a function of the area, the number of edge pixels, and the number of pixels on diagonal edges. Human hair, plastic line, and root samples, with widths from 0.13 to 1.65 mm and lengths near 100 cm, were measured manually and with image analysis. Mean lengths of four repeated scans after reorientation were within 0.4 to 1.2% of manually determined lengths. Four repeated scans (118 pixels cm-1) of reoriented roots yielded length estimates with C.V.'s from 0.31 to 2.18% on low density samples (< 500 cm). At higher densities, increased frequency of adjoining edges resulted in underestimated lengths and overestimated widths. A correction for these inaccuracies, based on the determination of mean width on a low density subsample, provided accurate root length estimates up to 10 m. Overall, this method overcomes limitations of other procedures such as (i) the requirement for random orientation of the sample necessary with the line-intersect method, (ii) reliance on a comparison to a standard curve from non-root objects, and (iii) inadequate imaging resolution.