JANUARY 2020 • COUNTRY LIFE IN BC Deep learning helps root out weeds


Tech points way to stronger


crops and fewer pesticides


by MYRNA STARK LEADER KELOWNA – Deep learning


and hyperspectral imaging is giving Mohsen Mesgaran a glimpse of the future of top- quality seed selection and weed detection. Mesgaran is an assistant professor in the Department of Plant Sciences of the College of Agricultural and Environmental Sciences at the University of California- Davis. It’s unlikely, if not impossible, to look at a seed with your naked eye and know the quality or sex of the plant it will produce, or if it’s a seed that will geminate well. But in his recent research, Mesgaran has demonstrated that both are possible using deep learning.


Deep learning is a


machine-learning technique. By analyzing masses of image, text or sound data, a machine uses neural networks and a series of filters to process information and extract features and patterns, like those in our brain, to identify and classify things. After seeing 900 to 1,000 pictures of a dog, for example, the machine can look at a photo it hasn’t seen before and identify whether or not it’s a dog. According to a video at


Mathworks.com, the technique has been around for some time and used by the post office to recognize hand-written postal codes and sort mail. But Mesgaran is working to apply deep learning to weed science and agriculture.


During a plenary session at


the Canadian Weed Science Society conference in Kelowna, November 20-21, Mesgaran shared that studies have proven that a computer with deep learning is now more accurate than a person. A key reason is the machine can access and store more data than any one person could ever know. The falling price of data storage and increase in computational power underpins the technology and has opened up applications in agriculture, he explained. In his research, Mesgaran is


proving ways deep learning can be applied. He showed how a variety of types of images of seeds from two different locations, as well as from water-stressed and non- stressed plants were pretty accurately classified. By looking at images of each


Mohsen Mesgaran says agriculture needs to incorporate state-of-the-art deep learning to help manage weeds in crops. SUBMITTED PHOTO


seed for things like colour wavelengths or patterns, deep learning was able to search for similar characteristics and determine from which of the two locations the seed originated with 85% accuracy. Water-stressed seeds germinated better, making them more desirable. He’s also been successful determining the sex of a plant from only its seed, eliminating the need for genetic testing or growing the plant. He shared examples of


research where deep learning counted the fruit on a plant


PTO GENERATORS


with 80-83% accuracy, detected plant disease, produced land cover classification maps with 85% accuracy and predicted crop yield.


While agriculture is among


the least-digitized industries, investment in ag tech is growing and Mesgaran says deep learning could help better target or change pesticide use. “Weeds are smart. They


have many hidden layers with millions of parameters, and we can only outsmart them using AI [artificial intelligence]


rather than ai [active ingredient],” quipped Mesgaran.


Specific to weeds, there is identification of weed type, density and weed growth stage. Each can be identified using semantic segmentation, which is the process


computers use to identify and classify each pixel of data stored in a digital image. And, because deep learning can detect weeds under varying environmental


conditions, it has huge potential in site-specific weed management, allowing herbicide application to individual plants at precisely the right stage of growth. “In weed research and management, deep learning offers a new set of valuable implications from accurate discrimination of weeds from the crop and automated weed identification to examining the seed viability or seed classification,” he said.


Tree Fruit Competitiveness Fund


The Province of BC has provided funding to enhance the competitiveness of the tree fruit sector.


The fund is open to tree fruit growers, producers, and processors to support three key areas of priority: ● Research: cultivar, disease and pest research. ● Marketing: export market opportunities and market development research.


● Infrastructure: sector-based infrastructure modernization such as new equipment.


The Tree Fruit Competitiveness Fund is jointly delivered by the BC Fruit Growers’ Association and Investment Agriculture Foundation BC.


Pre-order your Baumalight generator now for delivery in 8 weeks and get an 8% discount.


For details about the Tree Fruit Competitiveness Fund, including eligibility and application forms, please visit www.bcfga.com or iafbc.ca/tree-fruit, or contact funding@iafbc.ca.


Project intake is continous. Apply in advance of project initiation – 8 weeks minimum is recommended.


Locate A Dealer Online


1-866-820-7603 | BAUMALIGHT.COM Dale Howe | 403-462-1975 | dale@baumalight.com


MFG A VARIETY OF ATTACHMENTS BRUSH MULCHERS | BOOM MOWERS STUMP GRINDERS | TREE SAWS & SHEARS


TREE SPADES | ROTARY BRUSH CUTTERS | TRENCHERS AUGER DRIVES | PTO GENERATORS | DRAINAGE PLOWS


BC FRUIT GROWERS’ ASSOCIATION 1-800-619-9022 (ext 1) email: replant@bcfga.com www.bcfga.com


33


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48