Wagengingen University is at the forefront of innovative solutions for the food industry – and the fresh produce sector is one of the main points of interests for those working in the university’s research institutes. Produce Business UK was on hand at the recent Amsterdam Produce Show to hear about a host of futuristic developments in the Netherlands.
One of the first rules of business is that you should never stop learning. Those who fail to prepare for the future or aren’t nimble enough to adapt to ever-changing market conditions risk being left behind. The fresh produce industry often has long established successful business concepts, however the best companies are those always looking to the future.
Robotics might appear to be something drawn from a dystopian sci-fi film – the fear of humans being replaced completely by machines and artificial intelligence – but when applied correctly, using robotics, innovative computer programmes and automation can assist everyone in the supply chain.
This was the premise of an engaging seminar at the recent Amsterdam Produce Show. Presented by Rick van de Zedde, senior researcher and business developer for Computer Vision at the Netherlands’ Wageningen University & Research. His presentation shared examples from the school’s research work in an effort to demonstrate how the ‘future’ is actually happening now in fresh produce circles.
“We are working on ways to automise particular tasks in horticulture,” Van de Zedde noted. “We are working on ways to automatically screen how tomatoes grow in greenhouses. We work on the automation and understanding how young, small Arabidopsis plants grow in climate rooms.”
Van de Zedde outlined how the research teams were working with drones to measure how plants grow in the fields and they were also working in retail on automated checkouts. However, the main thrust of his talk was concerned with fresh produce.
He explained: “We are aiming for a full understanding of what is going on in horticulture, in food production. It starts on a basic level. And then we go down to greenhouses, then down to fields, post-harvest techniques and then to the market. We are aiming for sensors and robots to understand quality.”
The title of his talk was Phenomenal phenomics: New ways to determine quality, and this required some unpacking. Phenomics, he explained, is concerned with measuring how a plant or product grows. How it is changed in relation to its genotype and its innate profile in relation to the environmental influences.
“We need to gain significant progress in improving the quality and productivity of crops,” Van de Zedde said. “We need to produce more food in this world. Therefore we need to understand and exploit what variety we’re talking about, so what genotype and phenotype of plants/products in a changing environment – taking into account climate change and/or diseases for instance. And how do they grow in these conditions? Can you find the right variety that grows optimally even in the harshest of conditions – for instance very hot summers in India.”
He took mangoes as an example. He noted that from the outset – the fruit growing on the tree – there is already variability. As the product is packed and shipped the quality is not improving. And so by the time the fruit is presented in supermarkets again there is marked variability. At Wageningen they have been trying to understand how this quality develops along the supply chain.
“This model,” Van de Zedde told the audience, “helps you with food loss prevention or waste management.”
Van de Zedde explained how they go about measuring this vital ripening process to get efficient and accurate data.
“We like to measure the initial ripeness of the product,” he said. “We like to understand quality development. We’re introducing sensors and automation in there. And we’d like to apply customised ripening protocols. Because what we say is quality is developing in the whole chain. We’d like to understand where it starts – what were the chain conditions? What were the quality developments? We then take all of this into account and place it in a model. From there we would like to predict the quality as soon as you measure a batch of mangos or a bunch of tomatoes somewhere. How is the quality going to develop in the forthcoming weeks?”
Firmness and ripening is measured using non-destructive sensors – the beginnings of automation. Van de Zedde hypothesised that correlations from measurements of avocados for instance could be drawn to other products.
They are also looking at batches of fruit going around the world. The units in which products are shipped are often in crates. Often these fresh products do not leave the crate before they enter the supermarkets, and so on crate level they have introduced a variety of measurement systems. From focussing on the shape with 3D sensors to examining colour to measure brown spots on the skin, they take into account a variety of parameters to model how the produce develops.
“We are working on crate level inspection methods,” he noted. “This means we are not measuring all products. We’re going to sub-sample big batches of produce, by selecting a few crates, measuring the produce inside those crates. This measurement represents the quality of the whole batch. So we’re focused on automation of current manual work. This work is often being done by human experts. We would like to bridge the gap between the lab inside companies and the current process with automation.”
Van de Zedde then pointed to a robotic arm that they have been utilising at Wageningen. He said work on automation has produced some excellent results on tomatoes – measuring the brix value of a tomato in an non-destructive and extremely fast manner.
“We are using an optical measurement device, and within 20 milliseconds we can measure the tomato and give feedback about the brix quality. Instead of destroying the tomato and squeezing it on a brix sensor. We can do this extremely fast – the sensor can be mounted on a robotic arm that can bring the sensor to the crate and work in specific spots.”
Robotics and automation, he posited, can also be applied within packaging. At present he outlined a lot of manual labour is used to fill a variety of packaging sizes, because humans are flexible.
“We know that manual labour is expensive but humans are flexible,” he said. “So there is a preference for using humans in this case, but we are focussing on using automation in there. But robots at this point are not flexible.”
This, he said, was the starting point for a big EU project called PicknPack.
There is a working demonstration at Wageningen, which shows a new way of working with fresh produce. Starting at the harvest crate all the way to retail.
Utilising this system users can look at the origin of the crate, the produce is unpacked and checked for quality.
“We’re using a lot of sensors to estimate immediately the quality of the produce. It’s screened with hyper spectral imaging – we can measure the brix, we can see defects immediately. This is currently a task that is being done by hand. We can take this information into account and give feedback including information for the consumer on the package. We can print information on the package.”
And then there is the automated checkout. EasyFlow is an automated checkout – Wageningen developed the brain for the checkout – that doesn’t rely on barcodes. The device can autonomously recognise 30,000 different products. Tiny sensors hidden under the belt recognise the products, the consumer pays and can leave the shop.
“We use machine learning in there,” Van de Zedde said. “And we are combining a lot of different sensors to estimate what is actually on the belt. Using computer vision, NIR technology, the weight of the product and advanced statistics.
“In a supermarket products are continuously changed. Packaging continuously changes, sizes are changed, colours are changed, branding is changed so this machine needs to be continuously self-learning. So what one checkout is learning it is sharing it with other checkouts.”
Concluding, Van de Zedde noted the need to collect data.
“It’s quite a complicated world. We need to measure a lot, we need to introduce a lot of automation to understand all the data streams and combine it in such a way to really optimise the logistic chain. We can collect and we can link data in the chain using automation. So we need to define and exploit the relation between non-destructive robot measurements and manual destructive lab measurements.”
What is absolutely paramount, he explained, was the need to understand and predict the development of quality in fresh produce. Everything is meant to mimic the human expert.
So automation is happening already in the fresh produce industry – certainly at a research level. How far will industry follow?