A method for estimating root volume and total root length using images of soybean root in hydroculture taken by digital camera

Studying the growth of soybean roots in hy- droponic culture provides a means to avoid issues caused by the heterogeneity and opaqueness of soil. In this study, we applied three salt-tolerant soybean near isogenic lines (NILs) and three corresponding salt-sensitive NILs for hydroculture with ten different salt concentrations. A digital camera was utilized to continuously capture RGB side-view images of hydro- ponically grown soybean roots. To extract two-dimen- sional information of root area from the side-view im- ages, we applied a machine learning-based method, the trainable WEKA segmentation. The root volume was estimated by raising the root area to the power of 3/2 and applying a coefficient. Additionally, skeleton length of root was obtained from the bottom 30% of images with minimal root overlap to calculate the av - erage root diameter. The total root length was esti - mated from the root volume and the average root di - ameter then applying a coefficient. The results indicate that the coefficients of root volume and total root length ranged from 0.0164 to 0.0291 and 0.0449 to 0.0874, respectively, for the three correspondent NILs. Using these coefficients, the root volume and total root length were estimated with the root mean square error (RMSE) ranging from 1.87 to 5.32 cm 3 and 0.67 to 2.32 × 103 cm, respectively. The estimation results clearly demonstrated the differences across NILs and treatments even though ten salt treatments caused different root morphologies. This method may contrib- ute to further understanding the dynamic changes in root traits under various abiotic stresses.