Over the last decade, our researchers have been developing a tool to more effectively and precisely assess quality in agricultural commodities. Neural networks is the best solution to this challenging problem.
The DuPont™ Acurum™ decision engine has been developed using actual seeds and Canadian Grain Commission grading factor definitions. DuPont is willing to work with local regulatory agencies to develop Acurum™ in areas outside Canada.
Acurum™ performs over 300 measurements of each kernel in a sample to measure certain visual factors.
How Artificial Neural Networks Work
Artificial neural networks (ANN) is a type of artificial intelligence that is based on the design and function of the brain. ANNs are able to extract meaningful results from complex, non-linear data that is often encountered in natural systems.
In ways that are analogous to the function of the brain, ANNs learn through example. In the case of classification problems associated with seeds of various types and conditions, ANN models are “trained” against a known type of seed.
Before being released to industry, ANNs for Acurum™ are “trained” using the following process:
- Seed sets of a certain type and condition (e.g., fusarium damaged wheat) are obtained from a large number of samples representing the growing region of interest.
- Images of the seeds are captured using the Acurum instrument.
- Data from seed images are extracted for a number of dimensional and colour variables.
- Image data is presented to the ANN repeatedly.
- The ANN model with the best and most consistent results is selected, tested and deployed to Acurum™ instruments.
ANNs are good at solving problems that are too complex for conventional technologies and are often well suited to problems that people are good at solving. However, the benefit of Acurum™ ANNs is that they can approach the discriminatory power of a human and do this consistently time after time on the same instrument and between instruments.
