A machine vision platform for measuring imbibition of maize kernels: quantification of genetic effects and correlations with germination

• Main Authors: Nathan D. Miller, Scott C. Stelpflug, Shawn M. Kaeppler, Edgar P. Spalding
• Format: Article
• Language: English
• Published: BMC 2018-12-01
• Series: Plant Methods
• Subjects: Germination; High-throughput phenotyping; Imbibition; Image analysis; Machine vision; Zea mays
• Online Access: http://link.springer.com/article/10.1186/s13007-018-0383-7

Abstract Background Imbibition (uptake of water by a dry seed) initiates the germination process. An automated method for quantifying imbibition would enable research on the genetic elements that influence the underlying hydraulic and biochemical processes. In the case of crop research, a high throughput imbibition assay could be used to investigate seed quality topics or to improve yield by selecting varieties with superior germination characteristics. Results An electronic force transducer measured imbibition of single maize kernels with very high resolution but low throughput. An image analysis method was devised to achieve high throughput and sufficient resolution. A transparent fixture held 90 maize kernels in contact with water on the imaging window of a flatbed document scanner that produced an image of the kernels automatically every 10 min for 22 h. Custom image analysis software measured the area A of each indexed kernel in each image to produce imbibition time courses. The ultimate change in area (ΔA) ranged from 19.3 to 23.4% in a population of 72 hybrids derived from 9 inbred parents. Kernel area as a function of time was fit well by $$A\left( t \right) = A_{f} \left( {1 - e^{ - kt} } \right)$$ At=Af1-e-kt where A f is the final kernel area. The swelling coefficient, k, ranged from 0.098 to 0.159 h−1 across the genotypes. The full diallel structure of the population enabled maternal genotype effects to be assessed. In a separate experiment, measurements of kernels of the same 25 inbreds produced in three different years demonstrated that production and storage variables affected imbibition much less than genotype. In a third experiment, measurements of 30 diverse inbred lines showed that k varied inversely with germination time (r = − 0.7) and directly with germination percentage (r = 0.7). Conclusions Nonspecialized imaging hardware and custom analysis software running on public cyber infrastructure form a low-cost platform for measuring seed imbibition with high resolution and throughput. We measured imbibition of thousands of kernels to determine that genotype influenced imbibition of maize kernels much more than seed production and storage environments. In some hybrids, k depended on which inbred parent was maternal. Quantitative relationships between k and germination traits were discovered.