Bosch uses AI to spot black-grass in fields

Bosch and its partners created a comprehensive and innovative technology package, where high-tech cameras mounted to a sprayer boom can detect black-grass (Alopecurus myosuroides) in cereal crops and the individual growth stage of each weed.  

This unique, precision technology ensures the right amount of herbicide is spot-applied to the specific field area, the researchers explained.

Two also highlighted the positive outcomes delivered by this technology. Firstly, farmers can save money because herbicide use is optimised via spot application and is not uniformly used across an entire field.  

Secondly, black-grass becomes less of a problem, and over time its prevalence is reduced. The core aspect of this project was the use of AI, which determined where black-grass is growing and the developmental stage of each weed. 

‘Truly collaborative project’ 

A variety of steps were needed to complete the project. Its initial objective was to see if black-grass could be identified. Rothamsted Research in a laboratory environment helped train Bosch cameras to detect black-grass by photographing it from a certain height over a lengthy period of time as it grew.  

Bosch then trained an AI algorithm to apply this knowledge to images that came from cameras that were driven across a field. The aim was to identify areas where the black-grass is growing. An important point is that the camera does not request the sprayer to target a particular plant. Instead, it asks the sprayer to target the area where it can see black-grass.  

As the project progressed, the setup of the Chafer Machinery crop sprayer was adapted, for example, with a greater number of cameras and different boom heights on each side of the sprayer. This continuous development eventually resulted in a fully equipped boom with 28 cameras.

Peter Frankland, application engineer at Bosch, said: “This was a truly collaborative project. The agronomic part of the process was managed by BASF Digital Farming, using its xarvio Digital Farming Solutions platform to generate customised maps for use by the Chafer sprayer and to determine the herbicide dose and type.  

“Rothamsted gave us the understanding of types of cultivation that farmers use and some methods of solving the black-grass problem without using crop protection. We worked together to ensure the best possible outcome, and we tackled the various challenges that came up.  

“For example, in our discussions about the impact of evolving farming techniques with Rothamsted Research, we learnt that farmers increasingly adopt no-till or low-till methods, and so weed control becomes a more significant challenge.” 

Mr Frankland added that unlike ploughing, which buries black-grass seeds and prevents germination, these practices leave seeds closer to the surface. “There’s a lot of science behind the process,” he concluded.   

Rothamsted Research had an additional way of confirming the project trial results. Its researchers used quadrants, precisely localised, to count black-grass occurrences. Each quadrant is consistently measured to check that the black-grass is reducing. One of the more challenging aspects of the project was to train the AI.  

Muhammad Kassem, AI expert at Bosch, explained how it was done: “At the start we didn’t have anything to benchmark this against. As we progressed through the project, the model became much more mature, to the point where we could detect black-grass on unseen images with a high level of accuracy.  

“Overall, we scanned around 5000 images, covering a variety of different seasons and where black-grass is growing among different crops, not only wheat but also barley. Each time the sprayer completed a pass across the field, we generated such a high volume of images that it took a couple of days to download them all.  

“Then we used coding to clean up the images, another code to convert them to the right file format and a final code to train the AI algorithm. It was a fantastic project, and we’ve come out of it with good accuracy, and that’s the key measure.” 

‘Enabling key innovations for the UK market’ 

Bharath Jayakumar, director of Global Key Accounts at Bosch UK, explained that the project started in 2021. 

He said: “Initially we scouted for suitable consortium partners, and together we then applied to Innovate UK. We secured funding from DEFRA and Innovate UK to develop a local use case for the ONE SMART SPRAY weed management system developed by Bosch and BASF Digital Farming.  

“As a small UK team, we’ve worked with Rothamsted Research, Chafer Machinery and BASF Digital Farming to validate the technology to scan and identify black-grass in wheat.  

“Peter’s expertise in machine integration was crucial here, and as the AI expert on the project, Muhammad analysed, cleaned and annotated the scanned images and trained the machine learning model to identify weeds among crops.” 

The scientists reached a model accuracy that enabled black-grass control and reduction over time. “This project demonstrates the opportunities we have in the UK to access government funding to enable key innovations for the UK market. It also supports the development of our associates, and the project fits with the objective of the Bosch Mobility UK strategy to grow in markets and sectors outside of our core business areas,” Mr Jayakumar concluded. 

Bosch was the lead partner in the research and development project consortium, which comprised recognised experts from BASF/xarvio Digital Farming Solutions, Chafer Machinery, and Rothamsted Research.  

Together they were awarded a grant of £1,452,614 for the three-year project from DEFRA and Innovate UK, the UK’s innovation agency, under the Farming Innovation Programme – Small R&D Partnership Projects. 

Funding for the black-grass project was awarded as part of DEFRA’s Farming Innovation Programme and the UKRI Transforming Food Production challenge. 

READ MORE: Fresh approach to evolving black-grass situations

READ MORE: How to manage black-grass when returning land to cropping from SFI options 

Read more arable news.