Automatic and Smart Monitoring Platform for Rice Seed Germination Based on Time-Series Images

Traditional seed germination experiments mainly rely on manual operation, which suffers from issues such as low efficiency, strong subjectivity, and susceptibility to damaging seedlings. To more efficiently and accurately perform quantitative measurements of phenotypic characteristic dimensions of seeds during the germination process and statistically analyze germination-related indicators, this study designed an automated seed germination monitoring platform, combining machine vision and deep learning technologies, to achieve intelligent monitoring and quantitative analysis of rice seed germination processes. The hardware system is based on an artificial climate box, integrated with a circulating lifting device and an HD camera module, supporting timed automatic acquisition of germination images. The software system integrates One-Class Support Vector Machine (OCSVM) and Faster R-CNN models for measurement and extraction of dynamic phenotypic features and high-precision classification, respectively. Experimental results demonstrate that the system can accurately measure morphological feature changes such as seed area, perimeter, convex hull area and radicle length. The Pearson correlation coefficient between the germination rate detection results of the two models and manual evaluation results exceeds 0.996. Furthermore, the system provides multi-dimensional data support for seed vigor assessment through dynamic analysis of radicle length and statistics on cumulative germination curves. Compared with other commercial equipment, the equipment platform in this study offers advantages of miniaturization, automation, and high flexibility, enabling the simultaneous conduct of multiple sets of seed germination experiments and more effectively meeting the practical needs of small and medium-sized enterprises. It provides a technical reference for the intelligent transformation of seed quality detection.