Digital image analysis for automated seed germination assessment in genebank viability testing

Genebanks play a vital role in conserving the world’s plant genetic resources. However, the capacity of seeds to germinate declines over time and culminates in the loss of viability even under the best long-term storage conditions. Genebanks monitor their seed collections regularly via germination assays to ensure that they remain viable, but methods can be laborious and subjective. The development of automated methods for assessing seed viability, using computer vision, offers the potential to remove subjectivity and improve accuracy. An experiment was established to test the application of image analysis techniques for quantifying seed numbers and measuring the viability status of seeds grown in Petri dishes under standardised germination protocols following storage. Seeds from five species of pulses and cereals were grown in 90 mm Petri dishes on black filter paper and manually assessed for viability at regular intervals. Measurement values for manual and automated detections of seed numbers and their germination status exhibited correlations as high as 0.98. Our algorithm was able to reliably detect the orientation of radicles and attribute viability to the correct seed. We envisage that this method, using image analysis, could be further developed to facilitate the high throughput monitoring of seed germination in genebanks.