An Automatic Method for Rice Seed Vigor Classification Via Radicle Emergence Testing Using Image-Processing, Curve-Fitting and Clustering Methods
Abstract Background: Rice seed vigor classification is important for seed storage management by seed producers and by farmers while planning their cultivation activities. Field emergence is a direct method of seed vigor testing but is laborious, time-consuming and subjective. The accelerated aging (AA) test is often used as an indirect method for rice seed vigor classification in the laboratory. However, the results from this method are often imprecise. This paper presents the SVRice package, a simple, cost-efficient and flexible procedure that utilizes computer image analysis for high-throughput, automatic rice seed vigor classification. SVRice consists of 4 steps: dynamic imaging, image processing, curve fitting and clustering. Seed vigor was classified based on radicle emergence indices, such as maximum radicle emergence (MaxRE), mean radicle emergence time (MRET), radicle emergence speed (t50), uniformity of radicle emergence (U7525), and area under the curve of the radicle emergence fitted curve (AUC).Results: Parameters used to classify rice seed vigor, such as MRET, U7525 and t50, were strongly negatively correlated with the saturated salt accelerated aging (SSAA) test. A germination time of 90 hours at 25°C was sufficient for effective classification based on SVRice, whereas the SSAA test took approximately 400 hours to complete. The SVRice software algorithm was created to be especially suitable for assessment after 6 months under controlled atmosphere storage (at 15°C and 37% RH in a hermetic bag). The study showed that SVRice could unambiguously classify 40 indica rice samples with different varieties, production years, production sites, storage times and storage conditions compared with the SSAA test.Conclusions: This paper confirmed the accuracy, reproducibility and flexibility of the SVRice package for automatic seed vigor classification of Oryza sativa seeds; moreover, it is also likely applicable to other species as a viable alternative to current methods that require more time and are less precise.