Development and demonstration of an automated, selective broccoli harvester


SummaryThe development of a fully automated harvesting system is crucial for broccoli growers who currently harvest manually and face the full impact of the National Living Wage by 2021. An automated harvester would address the urgent need to reduce labour and improve growers’ ability to control production costs. KMS Projects has spent £250k+ and 10 years investigating various options for an automated harvester resulting in a solution that combines imaging equipment with robotics and specialised software. A demonstration of their prototype successfully identifying harvest-ready broccoli in the field and directing a robotic arm into position ready to cut the head was given in 2015.

During 2015, a team from The  University of Lincoln successfully developed an alternative imaging system using low cost 3D Kinect cameras to accurately identify broccoli heads.  Their project aimed at developing the vision systems to also drive a robotic harvester. They are now looking to develop  RTK-GPS tools to map the precise location and size of each head across an entire field. The two teams plan to work collaboratively to build on their work to date to deliver a prototype which can work for long periods to cut, lift and collect broccoli heads of the preferred size, leaving unsuitable heads behind. This project will deliver a tried and tested prototype of a single module rig which can be quickly developed into a first-off commercial machine.

Project code:
CP 153A
01 June 2016 - 15 December 2017
AHDB Horticulture
AHDB sector cost:
Project leader:


CP 153a PhD Studentship - Annual Report 2018 CP 153a PhD Studentship Annual Report 2019

About this project

AimsA successful automatic broccoli harvesting solution must be able to undertake all of the following tasks successfully whilst under continuous forward motion, at a commercially useful speed across a field of broccoli plants:

A.    Accurately identify broccoli plants in the field, in all light conditions, including at night.

B.    Accurately measure the size of each plant head and compare it against pre-agreed criteria in order to establish whether or not it is suitable for cutting.

C.    Obtain accurate data regarding the position of each broccoli head selected for harvesting.

D.    Despatch the robotic arm with pinpoint accuracy to the precise location of each broccoli head selected for cutting.

E.    Cut each selected head, leaving the stalk and any immature or unsuitable heads undamaged in the field.

F.    Lift each cut head without damage.

G.    Enable cut heads to be collected without damage for transportation to the processing facility.

H.    Work continuously (in all but the most severe weather conditions) for long periods of time, to be demonstrated by a successful 24-hour continuous trial.


Tasks A, B, C & D have already been achieved by the KMS Projects’ team using the first prototype design which was developed and tested during the 2015 harvesting season as a separate project (CP 153). Tasks A & B have also been achieved by The University of Lincoln team whose solution has been tested through 2015 in the UK and the team are currently gathering additional data in Spain.  The team is currently developing a GPS based solution to undertake task D.

This new project seeks to incorporate learnings from the work carried out to date by both KMS Projects and The University of Lincoln.  We believe that most robust solution to this challenge will be achieved by integrating both teams’ prior work. The teams aim to undertake a comprehensive comparison between the commercially available LiDAR (3-D laser scanning) based imaging system used by the KMS Projects team and the less expensive one developed by UoL using Kinect, a low -cost 3-D structured light camera.  The purpose of this is to ensure that the harvester is built using the most competitively priced imaging system that is able to deliver the required levels of accuracy, reliability and scalability.

The two teams have also taken a different approach to the way in which the precise location of each head is determined.  KMS Projects have developed their system using a combination of encoders which have shown to be successful in providing the robot with positioning data for the heads selected for cutting.  The UoL team are developing their system using RTK-GPS.  At this stage, the teams believe that the best solution will probably be a combination of the two systems.  They aim to test both and to compare their efficacy and to evaluate a combined approach in order to ensure that the commercial harvester incorporates the most accurate and reliable system. A redesigned and upgraded prototype will be used to rigorously test and assess the alternative crop identification and location systems developed independently by the two collaborators.  This will enable the joint team to select and adopt the best of each.

A key focus of the project will be to develop the prototype to undertake tasks E – H.  The rig will then undergo rigorous and comprehensive testing throughout the 2017 UK broccoli season. The overall aim of this project is to develop a single module rig which successfully automates the selective harvesting of broccoli in commercial environments where broccoli is grown in either 2 or 3 row planting schemes (between the tractor wheels).