Objective measurement for seed quality
GroundCover™ Issue: 119
State-of-the-art imaging technology is being used to speed up the identification and objective measurement of seed quality characteristics that guide pulse breeding.
The seed quality of pulses is primarily defined by physical characteristics such as size and colour. Generally, quality has been determined using sieving tests that are time-consuming and require a large sample. This means that the tests cannot be used until the later stages of the pulse breeding cycle, when sufficient grain is available. Other quality traits, such as seed coat colour or defects due to disease, weather or insect damage, are determined visually and are therefore highly subjective.
Delays in applying these tests due to insufficient seed can result in breeding lines with defective quality traits being carried through the program only to be discarded at a late stage.
The aim of this GRDC-funded project, ‘Objective high-throughput technologies for the pulse industry’, is to develop tools to speed up and objectively measure these quality traits.
Additional traits such as dehulling and splitting efficiency are also tested because most pulses are processed in this manner.
To develop high-throughput testing technologies that would allow germplasm screening to start at an earlier stage of the breeding program, researchers decided to explore the rapidly developing digital image technology.
For project leader Dr Joe Panozzo and researcher Linda McDonald, from the Victorian Department of Economic Development, Jobs, Transport and Resources, the project has involved developing mathematical models for each seed quality trait based on information extracted from images of the seeds.
Thousands of lentil, field pea, chickpea and faba bean seeds were painstakingly and individually characterised to determine the seed dimensions, volume and colour. This manually determined data was compared with the digital images and used to develop algorithms for assessing each quality trait.
In the process, images of up to 2000 seeds are captured using a high-speed camera in conjunction with a built-in laser, which records the surface height of the seeds as they travel along a conveyor. Once the images have been captured, processing software, developed as part of the project, enhances each image and identifies the boundary of each seed, measures colour variation to determine defects and applies the laser height contours to determine seed shape, volume and any damage.
The project uses germplasm from Pulse Breeding Australia (PBA) breeding programs – field peas, lentils, faba beans and chickpeas – to improve the accuracy of the models and to expand the range of traits that can be measured. Recent examples include the determination of seed volume and prediction of seed weight.
Earlier this year, digital image analysis was successfully implemented in the PBA field pea and lentil breeding programs, achieving more than 98 per cent accuracy in measuring seed size distribution and 100 per cent in seed weight. The analyses were completed within three weeks of harvest, which enabled breeders to select germplasm for desirable quality traits before sowing for the first time, providing a full season’s advantage over the usual seed-sieving methods.
For the 2015 harvest, the project team plans to install the equipment at a receival point, collect harvest data and compare the accuracy of the seed size data using image analysis to the data collected by the grain inspectors using the industry-standard sieves.
The long-term objective is for digital image technology to be adopted at grain receival, replacing the present subjective measurements of pulse quality attributes and defects. The goal is to develop an objective measure of seed quality traits that will provide a more precise assessment and deliver more accurate prices for growers.
More information:Dr Joe Panozzo
0418 365 310
GRDC Project Code DAV00132