Monitoring soybean's tolerance to flood stress using an image processing technique

Flooding is a common environmental stress that suppresses plant growth. Varietal differences in responses to flooding were generally reported as seed yield, but plant height, bio-mass, leaf color, and chlorophyll content also have been used. Manual measurement of these traits is tedious and often time-consuming. This study was conducted to test the use of machine vision extracted plant features for early, noncontact, and quantitative detection of plant responses to flooding. Top projected canopy area (TPCA) of soybean plants during the 2-wk flooding was obtained by an image processing technique. Plant movement was calculated from the average daily TPCA. Plant movement in response to flooding can be divided into two phases: the first "reactive" phase occurs immediately after the stress and is followed by the second "acclimatized" phase. During the reactive phase, flooded plants moved at greater amplitude in both positive and negative directions than control plants. During the acclimatized phase, flooded plants and control plants moved at similar amplitude. The duration of the reactive phase was 2 to 4 d shorter in the flood-tolerant "107" genotype than the flood-susceptible "157" genotype, indicating that the flood-tolerant genotype acclimatized more quickly to flooding stress than the flood-susceptible genotype. Within the first day of flooding, canopy growth detected by TPCA was reduced by about 50%. Regression analysis of the daily maximum TPCA values against day of flooding fits the two-straight-line spline models with R-square values from 0.82 to 0.98. The slope of the growth curve and the point at which the slope changes reveal the more flood-tolerant characteristics of genotype "107" as compared with genotype "157". TPCA offers a nondestructive technique, more sensitive than plant height, to quantify canopy growth and detect genotypic differences of soybean to flooding stress.