How emerging imaging technologies are helping the agri-food sector
Posted on 18/10/2017
By Charlie Winkworth-Smith, Knowledge Transfer Manager, Emerging Technologies
The agri-food sector faces many challenges, particularly the need to increase production to feed a growing population whilst limiting environmental impacts. The decreasing cost of cameras and sensors, accompanied by a huge increase in computing power means that new imaging technologies are now viable options for the agri-food sector. These emerging imaging technologies will add value across the sector, from optimising inputs using precision agriculture to increasing the efficiency of food manufacturing lines.
Imaging for arable farming
Today, there are more than 200 non-military satellites orbiting the earth and providing information to farmers. Satellite hyperspectral imaging has already made great strides in helping farmers get the most from their land whilst reducing inputs. By measuring the moisture content or nitrogen levels using remote sensing, water or fertilisers can be applied at variable rates across a field. One of the drawbacks of traditional satellite imaging, particularly in the UK, is frequent cloud cover. Synthetic aperture radar (SAR) can image through clouds and at night, but has until recently been prohibitively expensive for most users. However, as costs fall, SAR is beginning to become available for use in precision agriculture.
Unmanned aerial vehicles (UAVs or drones) are a great addition to the farmer’s tool kit. New imaging technologies such as lightweight RGB cameras and thermal imaging systems fitted to UAVs could soon be able to detect early signs of disease or pest infestations. The Food and Agriculture Organization of the UN (FAO) has begun preliminary field tests using drones to detect locusts and guide ground teams so the pests can be eliminated before they take flight and begin their destructive journeys.
When it comes time to harvest, new imaging technologies will transform labour intensive fruit and vegetable picking. Researchers at the University of Lincoln are developing a robot harvester that will be able to pick broccoli six times faster than humans and reduce the high levels of wastage that results from manual harvesting. The 3D vision system will assess the ripeness of the vegetable and decide whether to pick it or leave it in the field for another day, providing the farmer with incredibly granular data about the crop. This real-time information will help reduce wastage and improve the quality of the produce. Recent lab based trials have shown that similar 3D imaging systems can reduce pesticide usage by 72% by spot spraying weeds autonomously.
The Hands Free Hectare project led by a UK company, Precision Decisions, working with researchers at Harper Adams University aims to be the first in the world to plant, tend and harvest a crop using only autonomous vehicles. They have developed a driverless tractor that is able to follow pre-programmed routes around a field using global navigation satellite systems (GNSS). The addition of 3D vision systems has the potential to allow tractors to become truly autonomous, so that they have the superior perception required to navigate complex agricultural environments. By combining these technologies with infrared imaging, night time farming could become a real possibility in the future, which will have huge cost saving implications.
Post-harvest, new imaging technologies can help detect poor quality produce. Blackheart, which is an internal defect in potatoes that develops due to improper post-harvest storage, is currently detected using destructive tests that account for 0.5% of post-harvest wastage with a net value of more than £10 million a year in the UK. New automatic and non-destructive detection methods are currently being developed, such as visible/near infrared transmittance spectroscopy or microwave imaging, that can be used to identify spoilt potatoes and will help reduce waste and speed up the detection process.
Arable farming is not the only type of agriculture to be benefitting from advances in imaging technologies. Cainthus, an imaging company based in Ireland, has already made their first commercial sale of a machine vision system for identifying cows using facial recognition. The system identifies any animals that are not eating or have increased aggression and by measuring the arch of a cow’s back can give an early sign of a problem with its health. Researchers are now working on continuous monitoring systems that will be able to assess behavioural patterns and measure the animals’ health and welfare in real-time. Non-invasive imaging systems can also give a useful indication of fat and muscle mass before slaughter.
A major problem in salmon farming is the proliferation of sea lice. Traditional methods of eliminating them, such as adding other fish into the sea-cages to eat the lice or by introducing chemicals such as hydrogen peroxide, are never 100% effective. A new stereo vision system has been developed which uses cameras to locate the parasites on the salmon. The vision system then calculates the position and velocity of the fish and fires a high powered green laser to delouse the salmon whilst ensuring that the fish’s eyes are nowhere near the firing line. The cameras also help assess the behaviour and health of the fish. The system is already being used at 100 salmon farms in Norway and was introduced in Scotland in 2016.
Imaging in food processing and retail
Imaging technologies are helping improve quality and safety during food processing. Optical sorting and x-ray imaging are already widely used in the food industry to detect foreign bodies such as sticks, bones or glass. New imaging techniques are now being developed which will soon be able to rapidly detect food-borne pathogens. For example, a novel microscope using acousto-optic tunable filter based hyperspectral imaging which can rapidly detect pathogens such as Salmonella, Staphylococcus and E. coli has recently been developed by researchers at the Agricultural Research Service in the US. Hyperspectral imaging is also ideal for multicomponent analysis where the fat, sugar and moisture composition of a food product – such as a chocolate bar – can be mapped due to their individual spectral features. Integration of multiple techniques such as fluorescence and hyperspectral imaging achieves even better performance in detecting quality traits in meat products.
In retail, machine vision is now being used in some stores to identify food products without using bar codes. Unpacked fruit and vegetables can be identified by the supermarket cashier simply holding them up to a camera. Amazon is taking this even further by eliminating cashiers altogether. The first Amazon Go store was launched last year where shoppers scan themselves into the store using their mobile device and shop as normal. A plethora of cameras and sensors then track their every movement. Deep learning algorithms combine this imaging data with pattern recognition and weighing scales on the shelves to determine which product the customer has chosen. Once the shopper has finished, they simply walk out of the store and their Amazon account is charged.
KTN event on emerging imaging technologies in agri-food, 23rd January 2018
KTN ran an event in Birmingham on the 23rd January 2018 on emerging imaging technologies in agri-food to bring together the imaging and agri-food communities. You can read the report from this meeting here.
If you would like to find out more, please contact Charlie Winkworth-Smith at KTN to discuss further.
Like this story? You may also be interested in these articles on related themes: