Author: Anna Underhill. We all know that breeding new fruit varieties takes a long time. Read on to hear how graduate students at UMN are making it go a little faster.
As grape harvest winds to a close at the Horticultural Research Center, another season is ramping up – it’s data collection time!
Measuring physical fruit traits – like cluster weight and length, berry count, size, and color – is important, but it’s also tedious and time-consuming. This is a common problem throughout plant research, as scientists and breeders alike struggle to collect information on increasingly larger populations.
It’s this common problem that’s given rise to a trend many are embracing: a practice called "high-throughput phenotyping." Though it may sound convoluted, it’s as simple as finding ways to reduce hands-on work and collect data faster, all while improving accuracy. You’d be hard-pressed to find researchers who aren’t implementing high-throughput phenotyping somewhere in their pipelines, and the UMN Grape Breeding group is no different.
Disease problems in grapes are related to the shape and compactness of the clusters. Overly tight clusters lack airflow and form a favorable microclimate (or habitat) for fungi like Botrytis (grey mold) to grow. Measuring how tight the clusters are is important in choosing the best grape varieties.
However, measuring compactness is tough: It requires human judgement. It’s usually rated on a scale of 1 to 9 – and the standardized rating scale isn’t always appropriate for the hybrid grapes we work with. Berry color is another trait that requires judgement: what separates pink from grey, or green from yellow on a color spectrum?
Through image analysis, we’re able to use computers to measure these traits for us, saving both time and effort and increasing accuracy. Here’s how it works:
1) Photos of fruit are taken in the lab. These photos are then segmented into three components – berry, stem, and background – using a semi-automated process based on color. This color data can be recorded using any number of indexes (RGB, HSI, CIELAB) and used for analysis.
2) Next, researchers process the segmented images using a software program that measures important characteristics like cluster length, width, and proportion of each cluster that’s made up of berries, stem, and empty space.
3) Then, the researcher uses these measurements to calculate how compact the cluster is. The measurements are also useful on their own: for example, computer-measured cluster length only deviated 9mm on average from hand-measured cluster length. After all this, images can be catalogued and used as a visual record for each vine from year-to-year.
All of this leads to better fruit varieties, faster.
Author: Anna Underhill, Master of Science student, Grape Breeding and Enology
As grape harvest winds to a close at the Horticultural Research Center, another season is ramping up – it’s data collection time!
The Tedious Act of Fruit Research
 |
| Photographing grape clusters for faster and more accurate data collection. Photo: Anna Underhill |
It’s this common problem that’s given rise to a trend many are embracing: a practice called "high-throughput phenotyping." Though it may sound convoluted, it’s as simple as finding ways to reduce hands-on work and collect data faster, all while improving accuracy. You’d be hard-pressed to find researchers who aren’t implementing high-throughput phenotyping somewhere in their pipelines, and the UMN Grape Breeding group is no different.
Speeding up Plant Breeding
Image-based data collection is improving how important fruit traits like grape cluster compactness (tightness) and berry color are measured.Disease problems in grapes are related to the shape and compactness of the clusters. Overly tight clusters lack airflow and form a favorable microclimate (or habitat) for fungi like Botrytis (grey mold) to grow. Measuring how tight the clusters are is important in choosing the best grape varieties.
However, measuring compactness is tough: It requires human judgement. It’s usually rated on a scale of 1 to 9 – and the standardized rating scale isn’t always appropriate for the hybrid grapes we work with. Berry color is another trait that requires judgement: what separates pink from grey, or green from yellow on a color spectrum?
Through image analysis, we’re able to use computers to measure these traits for us, saving both time and effort and increasing accuracy. Here’s how it works:
1) Photos of fruit are taken in the lab. These photos are then segmented into three components – berry, stem, and background – using a semi-automated process based on color. This color data can be recorded using any number of indexes (RGB, HSI, CIELAB) and used for analysis.
2) Next, researchers process the segmented images using a software program that measures important characteristics like cluster length, width, and proportion of each cluster that’s made up of berries, stem, and empty space.
 |
| A computer program measures the shape, size, and tightness of the cluster. Image: Anna Underhill |
3) Then, the researcher uses these measurements to calculate how compact the cluster is. The measurements are also useful on their own: for example, computer-measured cluster length only deviated 9mm on average from hand-measured cluster length. After all this, images can be catalogued and used as a visual record for each vine from year-to-year.
The Impact on Minnesota Hardy Fruit
Using time and resources more efficiently is always a goal, and image analysis is one way we’re moving toward it. The benefits of high-throughput phenotyping don’t just offer a respite from long days spent measuring fruit in the lab – they offer a chance to carry improved accuracy further down the research and breeding pipeline. The improved accuracy can also make genetic mapping more precise.All of this leads to better fruit varieties, faster.
Author: Anna Underhill, Master of Science student, Grape Breeding and Enology