A new computer system under development at the Agriculture Canada Lethbridge Research Centre could improve the seed quality assessment process for the grain industry and create new marketing opportunities for Canadian wheat.
Scientists at the centre are developing a new computer-based seed analysis system that uses artificial intelligence to classify wheat seeds, as well as identify and quantify diseases and seed damaged by handling or the environment.
The development team includes Eric Kokko, an image analysis specialist, Bernie Hill, a neural network modeler, and Bob Conner, a cereal pathologist.
Read Also
Anti-separatist movement targets rural Alberta
Former deputy premier Thomas Lukaszuk’s anti-separatism Alberta Forever Canada petition campaign expects to run full steam ahead into the province’s farming regions
The new computer system can automate and standardize the seed quality assessment process at prairie elevators and coastal terminals. The system could also be used as a management and marketing tool to provide the Canadian wheat industry with an edge over its competition.
The imaging system analyzes groups of more than 1,000 seeds at a time, measuring physical characteristics such as length and width, as well as the color characteristics of the seeds.
This information is then processed using neural network modeling.
Neural networks are computer software programs patterned the way medical researchers believe the human brain works.
Researchers “train” the networks by inputting as many relevant examples of a given situation and as many variables as possible. The computer uses this data as experience, and when presented with a new situation, sifts through the data, mixing and matching the variables until it determines the most likely solution.
“The big difference between neural networks and the human brain is the human brain can deal with only three to four variables when making a decision, while the neural network can look at dozens of variables simultaneously,” said Hill.
In the seed assessment system, the data put into the neural network is the physical and color characteristics that identify different seeds. The computer system uses the data to characterize the seeds at the elevator, to identify and quantify diseases such as black point and fusarium head blight, handling defects such as broken or degermed seeds, and things such as bin burn or green seeds.
“There’s also potential to count frost-damaged seeds, sprouted seeds, as well as other factors such as contamination by weed seeds or insect frass,” Kokko added.
“Current seed quality assessment at elevators is a subjective and time-consuming process with an inherent degree of variability,” said Conner.
“The automated system is more objective, less variable, and will reduce the time required for seed assessment.”
Glenn Coulter, assistant director in charge of commercialization and marketing said: “We anticipate a lot of interest in this from the grain industry, including both grain handlers and processors.
“The industry is interested in rapid instrument technology to automate seed quality assessment anywhere from the farm to final domestic or international market destination.”
As an added benefit, the new seed assessment system could also help Canadian farmers compete in export markets.
“The challenge is to prove we can meet certain standards in export markets with this technology,” Coulter said.
“The seed assessment system could help Canada guarantee its customers they are receiving wheat fitting their specifications.”
The Lethbridge team expects to unveil the new grain assessment system this year.
