By Greta Chiu
There’s a wireless, self-powered imaging device the size of a credit card being tested in Alberta.
Known as “Firefly,” the IoT (Internet of Things) imager is being studied for use in early detection of disease and pest outbreaks among crops.
To do this, Alberta-based LEAN Systems and technology partner, Proxilogica, have teamed up with Agriculture and Agri-Food Canada at the Lethbridge Research Development Center (AAFC-LeRDC).
Dr. Anne Smith, a research scientist at LeRDC, and colleagues Drs. Jonathan Neilson and Charles Geddes have been developing inexpensive imaging platforms for digital image capture and image analysis of plant development in growth rooms, greenhouses and laboratories.
Applied in a commercial agricultural setting, the device can be used in virtually any environment and for any crop, with the exception of mushrooms as the device needs light to operate.
Bill Halina, managing director of LEAN Systems says two research groups at AAFC have started deployment in greenhouse and shelf propagation environments, with four more in the queue. “As we roll out more this summer we have also lined up some Lethbridge-area commercial greenhouse operations and we are open to interest from all segments of the market.”
Over the last two years, Smith and Neilson have been collaborating with LEAN Systems to test this technology. The early systems have been shown to effectively capture images of greenhouse plants, then upload them remotely to a central server. The images are then downloadable by the user. Image analysis protocols can be used to automatically extract information on plant growth over time.
In the meantime, the collaborators have started the first data capture campaign, collecting thousands of sensor readings and images of insect traps and plant structures. “We need this data to train the machine learning systems so they can discriminate between benign and pest insects and also identify disease signatures in leaves,” says Halina.
There are two ways in which the Firefly can support greenhouse producers: monitoring and analysis.
“In the first instance, arrays of Fireflies provide wide-coverage 24/7 scouting services looking for pest and disease outbreaks,” says Halina.” In this case, prompt reactions can prevent rapid spread and more widespread losses and so the system would be configured to emit email/SMS alerts to pest management staff. “
The second is as a preemptive decision-support tool. “Analytics software, with access to extensive historical measurements of all environmental readings and favourable/un-favourable outcomes as determined from the images, will be able to generate pre-emptive warnings when conditions that have led to issues in the past are detected in the current state of the greenhouse. For example, the machine learning system might determine that a combination of higher relative humidity, lower light levels and lower air speeds are correlated with mold outbreaks and therefore provide a ‘heatmap’ of where conditions should be adjusted to avoid such problem,” Halina explains. “Prevention is much better than cure.”
According to Dr. David Southwell, CEO of Proxilogica Corporation, behind the Firefly architecture are “large fleets of tiny imagers that maximize space coverage with enough on-board intelligence to pre-process and securely transmit data to the cloud where bird’s eye maps are then assembled. AI-boosted analytics functions may be performed at both edge and core, drastically reducing network traffic thereby enabling fleet scaling.”
Halina says it will be possible to directly interface with third party management software or closed-loop environmental control systems in the future. “We recognise that advanced greenhouses are already equipped with monitoring and control systems – these tend to be centralised. What we are bringing to the table is a complementary approach that emphasises fleets of tiny devices that can generate 2D (or even 3D in the case of vertical farms) maps of conditions, capturing variations and microclimates within a facility.”
The technology has applications in the field as well. As Southwell explains, “We start in CEA spaces and will soon be ruggedizing and adapting the imagers for more demanding open field horticulture applications, including UAV platforms.”
With files from: LEAN Systems