Tuesday, 10 July 2018
Could a fully automated broccoli harvester that saves growers significant amounts of time and money soon be a commercial reality? Sarah McWilliam, KMS Projects, updates on the AHDB-funded project aiming to do just that
Around 75,000 tonnes of broccoli are grown and harvested by hand each year in the UK. With the growing challenge of access to affordable labour, the development of a fully automated harvesting system is therefore crucial for UK growers.
Designing a selective, automated harvester to replace the manual system is no easy task. It must accurately identify and measure the size of each broccoli head and select only those which fit the grower’s pre-determined profile for picking. Then, each selected head must be cut and collected without damaging the uncut crop. All of this must be achieved in real time while the rig is in continuous motion.
After proving their concept in 2015, KMS Projects has been busy working on a follow-up project (CP 153a) to develop the concept into a single module rig capable of selectively harvesting commercially grown broccoli.
KMS Projects’ harvester is designed to be a stand-alone rig powered by and mounted on the front of a standard tractor. It uses sophisticated imaging and data analysis techniques to identify the broccoli plants and to select heads of the required size for harvesting. The cutting mechanism is moved swiftly into position by a six-axis robotic arm and the heads cut and lifted leaving the rest of the crop undamaged in the field.
The team built a prototype rig and tested it throughout the 2016 and 2017 growing seasons, refining and improving on various elements of its design before demonstrating it working at twice the speed of a manual cutter to an audience of invited guests in November 2017.
The demonstration was the culmination of a great deal of work, some of it genuinely ground-breaking, to overcome the challenges inherent in this project. Chief among these was to find a reliable method of accurately accounting for the forward motion of the rig over the field. This is vital so that the location data of the broccoli heads obtained by the imaging system could be used to send the robot to each head with pin-point accuracy without being affected by changes in tractor speed or the unevenness of the ground.
Another huge challenge was to find a way for the harvester to reliably identify broccoli curds partially hidden by leaf cover. The solution is perhaps the most exciting aspect of the project. The team have created a method of detection which uses ‘deep learning’ artificial intelligence techniques to train the rig to classify different sizes of broccoli head with impressive accuracy. One of the key advantages of this method is that the imaging system is continually learning from the data it collects.
The HMI (human machine interface) has been designed and work has already begun to integrate it with ISOBUS, taking the concept closer to a turn-key solution.
The rig is equipped with its own communications hub and a user-friendly website interface has been created which will provide access to performance analytics using real-time updates direct from the harvester both to the tractor driver and remotely sited farm managers.
The harvester collects location and size data about every broccoli head it passes over, including those not immediately selected for cutting. This real-time information can be used for yield mapping and also for improved harvest planning, as decisions will be informed by accurate, detailed data.
The end game
The team’s ultimate goal is to see their prototype developed into a three-module, commercially available machine. This machine will be capable of operating 24 hours a day, without tiring. It will cut broccoli heads that exactly match customers’ specifications, reducing growers’ reliance on manual labour and cutting harvesting costs significantly.
The harvester’s systems and processes which identify, locate and cut broccoli heads have obvious applications with other crops such as cabbage and lettuce varieties.
KMS Projects have a programme of R&D work planned which should deliver the first multi-module prototype machine during 2018. Assuming that operational tests of the prototype are successful, a commercially available multi-module machine should therefore be available to growers in the foreseeable future.
To find out more about the automated broccoli harvester, contact firstname.lastname@example.org.
AHDB’s SmartHort project is working to address the challenge of access to affordable labour through better management practices and smart technologies. To discuss your ideas about how the project could continue help UK horticulture, contact email@example.com