Cauliflower pickers must part leaves and feel each head by hand to judge if it is ready for harvest, a slow, skilled process that generates significant waste. The problem is that individual plants grow at different rates depending on local soil, moisture, and nutrients, and the developing curd is hidden beneath a leaf canopy, making accurate field-level forecasting nearly impossible. In 2019, the UK produced 82,000 tonnes of cauliflower but still needed to import between 10 and 30 percent of the crop to meet demand, while waste from missed harvest windows and quality errors remains high. This project aims to replace guesswork with data. Researchers will mount state-of-the-art cameras on farm equipment and use artificial intelligence to predict head size and growth rate for each individual plant. Combined with bespoke growth models, the system would tell growers exactly which fields, which sections of a field, or even which specific plants are ready for a retailer’s specifications. If it works, harvest teams could be deployed fewer times and target only the plants that are ready, reducing labour costs and cutting waste in the fresh-produce supply chain. The research is applied and commercial, with a direct route to reducing food waste and improving the efficiency of UK brassica farming.
View original technical description
Cauliflower is a staple of the British diet, playing a key role in a Sunday roast, Christmas dinner and cauliflower-cheese. In 2019 pre-covid, 82,000 tonnes were produced in the UK, but we typically need to import between 10-30% of the crop to satisfy demand. At the same time, there is a large amount of waste in cauliflower production worldwide, including in the UK. A key problem is that it can be difficult accurately to forecast growth of the cauliflower heads at a whole field level, and more particularly for individual plants. Different plants grow at different rates depending on local soil conditions, moisture and nutrient levels, and the growing curds are covered by a layer of leaves, making it impossible to see how big each head of cauliflower is. The problems translate to harvesting: for brassicas, harvest remains an extremely manual and high-skilled job, with teams of pickers deployed several times to each field selecting cauliflowers that are ready by parting the leaves to visualise the heads and then feeling them with their hands to assess size. Cauliflowers that fit the target size are cut with a knife and trimmed, but mistakes can be made, quality problems generate significant waste, and there can be over-production at inopportune times of year. Our concept is to introduce a step change in the way that cauliflower growth can be forecast. We aim to use state-of-the-art camera systems, coupled with artificial intelligence, to provide a method of predicting head size and growth rate for individual plants. Combined with bespoke growth models, this will provide the industry with substantially improved forecasting of the crop's readiness levels, thereby allowing pickers to be deployed fewer times and to target their harvesting operations to specific fields, specific regions of a field or even specific plants where the crop best fits a retailers' specifications.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know