Andrea Pufflerova of Photoneo explains how the UK’s National Physical Laboratory will use 3D machine vision in crop growing to feed future generations.
It is beyond dispute that one of the most pressing challenges the world faces today is to find ways to feed the ever-growing population in a sustainable way. Experts agree that crop yields will need to boost by 70 per cent to feed the population by 2070. Seed producers therefore endeavour to come up with the next variety of “Super Seed” that would tackle these concerns for future global food supply.
The traditional, manual method of capturing and analysing crop phenotype data is very time consuming and does not always provide all the information that is needed. Automation of this process therefore becomes absolutely necessary and presents a big step forward, opening up completely new possibilities for crop breeding programs. Wheat is the most widely grown crop in the world and one of the pillars of global food grain crop supply.
Currently, the average wheat yield is about 13 tons per acre but with the expanding population, this number will necessarily need to rise. The National Physical Laboratory in the UK has developed crop scanning technology using 3D machine vision to address these concerns. The technology aims to automate and improve the process of collecting and analysing crop phenotype data which is then used for breeding new crop varieties.
Data collection and analysis
Each year seed producers send their wheat seed samples to be graded. Based on the results, farmers decide which seeds to purchase for next year’s planting. To create a better variety of wheat, seed producers use phenotype data to guide their crossbreeding programs. The four basic factors that indicate the quality of wheat and how much a particular seed variety could yield are the ear length, ear height, volume, and the number of spikelets (grains per ear). This data is currently collected manually, using a ruler. However, this is very time consuming as the measurements need to be conducted across hundreds of field plots. The data might not be completely accurate as it is being taken from manual measurements and pictures. The 3D machine vision method developed by the National Physical Laboratory uses a 3D scanner and point cloud analysis and promises much better results and benefits in terms of time reduction and cost savings.
Automation takes the lead
The NPL uses a wheeled rig equipped with a number of 3D imaging techniques. One of them is structured light technology deployed in PhoXi 3D scanners. These scanners, produced by the Slovak company Photoneo and supplied by Multipix Imaging in the UK, are mounted on the rig construction to capture the scene within the frame in three dimensions, with a turnaround speed of 30 seconds. The standard sample area comprises two by five metres. The scanners provide point cloud information with millimetre resolution and incredible details of the individual grains. This information is then processed in MVTec’s Halcon imaging software with proprietary algorithms.
Continue reading article…