Partial automation of seed count tests

In seed testing, counting seeds from other plants helps to assess the quality of seed lots. The aim is to determine the possible presence of seeds from weeds and unwanted cultivated plants.

According to ISTA rules, the test is carried out by qualified analysts by visual examination of a sample of 25,000 seeds which are representative of the lot.

The aim of partially automating these analyses is to reduce the size of the sample to be visually analysed by at least 80%. Optical sorting, based on analysis of the physical characteristics of the seeds, separates the sample into 2 fractions. Using a camera and image processing based on artificial intelligence, unwanted or uncertain seeds are detected and ejected into a special bin. This fraction will be analysed visually by expert analysts.

The equipment was installed at the end of May and four laboratory staff have been trained and are currently working on developing and validating the tests.

The aim is to validate its use on a selection of species by the end of 2023.