Breeders who have developed a new tomato variety must register it before it can be introduced onto the market. For such a registration, the national examination office has to assess the new variety on a large number of characteristics. That is largely manual work. Within the European collaborative project INVITE, the Business Unit Greenhouse Horticulture and Flower Bulbs of Wageningen University & Research are investigating whether part of this can be automated by means of imaging and artificial intelligence (AI).
New varieties of flowers, plants, vegetables, and fruits are assessed according to the DUS method: is the new variety distinguishable from existing varieties (Distinctness), is the new variety and its possible fruits homogeneous (Uniformity), and are the character traits stable over consecutive seasons (stability). In order to establish whether a variety adheres to the DUS standard, bodies such as the Dutch Naktuinbouw score new varieties on dozens of characteristics. As scoring occurs manually and hundreds of varieties need to be checked for each crop each year, this procedure is highly labor-intensive. Within the European collaborative project INVITE, several research institutes are now looking for solutions to automate these procedures.
Measuring and weighing tomatoes
In the case of tomatoes, a registration involves 61 DUS traits, of which at least a third relates to the fruit. This means that, in order to fulfill a year’s worth of variety registration requests, experts have to manually measure, weigh, and assess thousands of tomatoes.
As partner institutes within INVITE, the Greenhouse Horticulture and Flower Bulbs Business Unit at WUR are working together with Naktuinbouw to develop a method for automating the description of new tomato varieties. The researchers have developed a prototype that employs a color and depth camera whose images are analyzed by AI to predict the trait scores as if they were assessed by crop experts.
Significant time savings
With this approach, the prototype is able to measure 3 DUS traits (such as size and shape). Towards the end of this project, this number will be expanded to at least 10 traits. Part of the description will therefore remain manual work. But if the most labor-intensive work is automated, this still means significant time savings, especially if, in the near future, the software can be used with a smartphone. Moreover, this AI-based approach makes it easier to standardize the working method at all the different European examination offices.
For more information:
Wageningen University & Research
www.wur.nl