Discriminating between seeds of ryegrass species using image analysis (BAGRASZAAD)

This project evaluates the feasibility of correctly assessing perennial and Italian ryegrass seeds using image analysis, particularly whether the certification standard (only 1% of other species present in a seed lot) is being met. Italian ryegrass seeds of variable size and morphology often contaminate seed lots of perennial ryegrass. The regulations for research related to quality control call for the independent and quantitative judgement of the purity of species and variety in each lot. The detection of impurities should be done as objectively and efficiently as possible. Because expert assessment of seed lots is very difficult and time consuming, we have tested the use of image analysis to determine whether it is an efficient and objective alternative for expert assessment.

For varieties of perennial and Italian perennial ryegrass differing in earliness and ploidy, 1000 seeds are analysed for shape, colour and texture using image analysis. We also conduct tests to see whether characteristics of the rachilla provide additional parameters for distinguishing between species. The most important parameters for classifying the species correctly are determined using discriminant analysis.

If the correct classification of species using image analysis is sufficiently high (>99%), a system can be developed to automatically evaluate ryegrass seed lots for the presence of different species. The goal is to reduce our dependency on experts by maximising the capacities of modern image processing. Difficult and time consuming activities (e.g. checking seed lots) can be performed in a more efficient and less labour-intensive manner.