Green-on-green spot spraying - nearing commercial reality?

Technological changes enabling rapid advancement for spot-spraying weeds


Crops
Aa

Next advance in spray technology is ability to spot- spray weeds in growing crop or pasture.

Aa
Cameras and processors situated every 3m on an Agrifac™ self-propelled sprayer. PHOTO AGRONOMO

Cameras and processors situated every 3m on an Agrifac™ self-propelled sprayer. PHOTO AGRONOMO

Key points

  • Advances in technology are allowing the pairing of cameras and artificial intelligence (machine vision) to identify objects in complex images
  • Machine vision will enable the spot spraying of weeds in fallow and crop, as opposed to current optical spot spraying technology of fallow spot spraying
  • Machine vision will enable advances in autonomous robots to conduct a range of tasks within paddocks

While green-on-brown optical spot spray systems for detecting and spraying weeds in fallows are now familiar to many growers (including WeedSeeker® and WEED-it™), the next big advance in spray technology is the ability to spot spray weeds within a growing crop or pasture, commonly known as green-on-green spot spraying.

Green-on-green spot spraying has been found to give the same reduction in herbicide and water usage found in green-on-brown technology, but does this in a crop or pasture.

The big breakthrough in green-on-green weed detection was the development of artificial intelligence (AI) and deep learning (a subset of machine learning), which has greatly improved the capabilities of machine/computer vision.

Green-on-green also offers the future possibilities of variable-rate applications of herbicides, fungicides and fertilisers.

Research at Wageningen University in the Netherlands has shown better growth rates and crop yields when green-on-green spot spraying has been used because more of the crop has not been exposed to the herbicide, which means the plant is expending less energy and other growth resources to detoxify (break down) the herbicide.

How green-on-green spot spraying of weeds works

The Bilberry™ system using cameras plus artificial intelligence sprays both green-on-brown and green-on-green, although the dry autumn in southern WA limited the options that could be demonstrated. PHOTO AGRONOMO

The Bilberry™ system using cameras plus artificial intelligence sprays both green-on-brown and green-on-green, although the dry autumn in southern WA limited the options that could be demonstrated. PHOTO AGRONOMO

The first research into green-on-green spot spraying of weeds in a crop or pasture relied on comparing colour differences between the weed and the crop, as well as leaf and plant shape.

The systems worked well in the laboratory but failed in the paddock. The failure at field scale was due to a range of factors such as variable lighting, bright sun versus overcast conditions, shading, movement of plants in wind and variable soil colour.

The big breakthrough in green-on-green weed detection was the development of artificial intelligence (AI) and deep learning (a subset of machine learning), which has greatly improved the capabilities of machine/computer vision.

A camera linked to AI is the set-up now most widely used for machine/computer vision to differentiate between complex images such as weeds in a crop. It is similar, in fact, to facial recognition technology.

Three technological changes enabled the rapid advancement of AI and deep learning:

  • improved computing power through the use of graphics processing units (GPUs);
  • huge sets of data ('big data'); and
  • powerful and complex algorithms through the stacking of neural networks.

Machine learning can be supervised or unsupervised. Unsupervised deep learning is something like a chatbot (a computer program able to hold conversations) that learns from its interactions with humans.

Bilberry™ chief executive Guillaume Jourdain outlined at GRDC Updates in early 2019 how the supervised process of deep learning is used for identifying wild radish in wheat with the following steps:

  • Define algorithm usage and objectives - recognise flowering radish in wheat with greater than 90 per cent accuracy.
  • Gather data - take in-field pictures of flowering wild radish in wheat under different conditions.
  • Sort and label data - on each picture, indicate what is wheat, wild radish, other weeds, stubble and so on. Also separate all images into two sets - training set and testing set. The training set is only used for training, while the testing set is only used for testing (the same image cannot be present in both sets).
  • Train the algorithm - show the training set thousands of times to the algorithm so that it learns the patterns.
  • Test the algorithm - show the test set (once) to the algorithm to compare the results of the algorithm with the reality.
  • Once happy with the results of the algorithms, go into the paddock to test. Paddock testing is the crucial part of the process.
  • Repeat until you reach your objectives.

Gathering enough images (data) in different situations such as bare soil, in wheat or in other crops, and with different stubble loads, is critical. Each image in the dataset must have everything in the image identified, which is labour-intensive and costly.

Guillaume Jourdain, Bilberry™, shows Craig White and Mitch Tuffley, Bayer, the finer aspects of the Bilberry™ camera and processing unit at Pingrup, WA. PHOTO AGRONOMO

Guillaume Jourdain, Bilberry™, shows Craig White and Mitch Tuffley, Bayer, the finer aspects of the Bilberry™ camera and processing unit at Pingrup, WA. PHOTO AGRONOMO

When will green-on-green spot spraying be available?

Several Australian universities are now at the early commercialisation stage for their vision systems.

John Deere/Blue River Technologies™ are currently operating their 'See & Spray' system on a limited basis in US cotton crops.

Bilberry™ has partnered with the Netherlands' Agrifac Machinery™ to use its system on Agrifac sprayers. Bilberry™ sees Australia as an exciting market for green-on-green spot spraying and plans to have calibrated algorithms available for in-crop weed spot spraying by 2021.

Current technology

In fallow weed management, green-on-brown optical spot-spray systems for detecting and spraying weeds are well established in Australian agriculture with the Weedseeker® and WEED-it™ technologies.

Both of these technologies have the following features. They:

  • measure the reflectance of green plants (using near infrared light), which triggers the nozzles to spray the weed;
  • can reduce the amount of herbicide used by 80 to 95 per cent, depending on weed density;
  • reduce the number of spray tank fills required (fewer litres of spray mix per hectare);
  • can be used day and night as they have their own light source; and
  • are now being used for selective cultivation (https://weedsmart.org.au/category/ask-an-expert) and targeting weeds with lasers and microwaves.

Some disadvantages include:

  • a high number of sensors (WeedSeeker™ has one per nozzle, while WEED-it™ has one per metre);
  • speeds are restricted to less than 20 kilometres per hour;
  • coverage can be affected by strong crosswinds, which move the spray pattern away from the targeted weed; and
  • a relatively high cost for the unit (up to $4000 per metre).

More information: GRDC Update paper, 'Green-on-green camera spraying - a game changer on our doorstep?'

Aa