Plant breeders try to develop new plants by crossing various species and evaluating and testing the resulting progeny. Much of the selection process is carried out by hand. Would computer selection make this any easier?
Researchers from Plant Research International have been working with colleagues from the James Hutton Institute in Scotland to explore the possibilities of monitoring capsicum plants in the greenhouse with computer image analysis.
Plant breeding involves observing huge numbers of plants to identify the genotypes that function best in a particular environment. Breeders also look for genes that will improve the flavour, shelf life or the yield of the plants. These genes can then be built into other genotypes to create better species. But this is a very labour-intensive and expensive process, so plants breeders are continually on the lookout for new and improved ways of measuring, characterising and selecting their plant material. For many years, all attention was focused on molecular techniques. But now that it has become possible to determine thousands of molecular markers at a relatively small cost, the breeders are facing another bottleneck: how to evaluate the plants as accurately and quickly as possible (phenotyping), and obtain insight into the relationship between the molecular markers and the phenotype.
The aim of the EU project ‘SPICY’ was to devise new methods and techniques for molecular plant breeding. Digital phenotyping using computer image analysis is one of these methods. When measuring plants with a camera, it is important that the camera has a good view of the plants. Setting up a camera system is difficult with plants like capsicum, which grow to a height of 3 metres, very close to each other on crop wires in a greenhouse. A special machine (the SPY-SEE) was developed for the SPICY EU project to enable researchers to monitor the development of plants in the greenhouse themselves. The SPY-SEE platform has 12 built-in cameras and moves between the rows of plants in the greenhouse. The images from the cameras are compiled into a 3D reconstruction of the plant. Details of this special method are described in Song et al (2011).
Computer image analysis enables the researchers to identify individual leaves in 3D, and track aspects such as growth and shape as they develop. The computer measurements can also be compared with the manual measurements, which shows a good correlation.
The computer measurements can also be linked to molecular markers to pinpoint the position on the chromosome of genes responsible for aspects such as plant growth. This is described in van der Heijden et al (2012).
This new method of monitoring plant features using computer image analysis will in future simplify the breeders’ task of evaluating their plants, and enable them to develop new species more accurately and efficiently.