Automatic assessment of variety traits delivers gains

News publication date: 07 December 2022

Variety research and testing requires a lot of assessment work. This is very time consuming and sometimes difficult. It would be nice if part of this process could be automated. There now seems to be some progress in that area. A result of cooperation between Wageningen UR and Naktuinbouw in a sub-project of the European INVITE programme.

INVITE

A new variety must be distinct, uniform and stable. To determine whether this is actually the case, Naktuinbouw performs DUS testing. These tests are time consuming. “Every year, some 200 to 250 applications are received for new tomato varieties. We assess the varieties for no fewer than 61 traits, says Wim Sangster, DUS specialist for vegetable crops at Naktuinbouw. “Some traits are fairly easy to assess and score, such as the colour of flowers or fruit. Others, such as weighing the fruit, take up a lot of time. Or traits can be difficult to standardise, for example a description of the shape of a fruit. In these areas, we could benefit from automatic and objective scoring methodology," he explains.

Vision technology

Naktuinbouw therefore started to cooperate in a project with technicians at Wageningen UR Greenhouse Horticulture (WUR). These technicians are specialised in rapid assessment using cameras and sensors (vision technology) and then interpreting the data obtained in this way. “The output had to be a tool that really boosted our methods. Something that could replace our visual observations in an efficient way. Even if it saves just one minute at each inspection, we can gain a lot. But if it takes a minute more each time, there’s no point or benefit for us," says Sangster. 

Three traits selected

WUR researcher Joseph Peller started work taking those wishes on board. “We first looked at which of the 61 scoring traits in tomatoes were suitable. It makes sense to automatically determine these traits if they are difficult to measure, require time consuming measurements, or if the trait is vital for determining the distinctness of a new variety. From the long list we ultimately selected three traits that met the criteria: fruit size, fruit shape ratios and the peduncle scar." In a laboratory environment there are effective methods that assess fruit using a number of cameras. “But most of the equipment is too large, too cumbersome and too expensive for use on location. So we performed the first tests with a RealSense stereo depth camera. But that proved to be unworkable in practice too. To prevent light causing a hindrance, we had to record the images under a blackout cloth. But that simply isn’t feasible," says Peller.

Simple solution

Fortunately, there is a much better camera available. It's relatively cheap and something practically everyone has on them: the smartphone. “You can record 3D images with a smartphone by moving it around the tomato in a certain way. And you can still record 3D images even if the tomatoes are in a crate using a technique called Simultaneous Localization And Mapping, or SLAM," says Peller.

Image recognition software is used to assess and score the images. This software can be installed on your smartphone via an app. This method satisfies the requirements: cheap, fast and accurate. Sangster is extremely curious to see if the method works under practical conditions, especially if the tomatoes are still in a crate. In that case, the work load would be significantly reduced. “If the method works effectively and delivers time gains, or greater objectivity, then we can apply it throughout Europe. And you can potentially provide hard evidence that a variety is genuinely distinctive, if the CPVO (the Community Plant Variety Office of the EU) asks for it. Automatic image analysis can be more objective than human assessment."

Perennial ryegrass

Peller and his colleagues are developing a method to assess the traits of perennial ryegrass for colleague organisation NAK. “For example, the height and the start of flowering. We are investigating whether we can capture those traits with cameras mounted on drones, and then analyse them correctly,” he says.

Major European project

INVITE stands for ‘INnovations in plant VarIety Testing in Europe’. INVITE is a five-year, European Union funded project that aims to innovate and improve plant variety testing (DUS and VCU). The main goal of INVITE is to improve the efficiency of the assessments, mainly through the application of new technology. This concerns the introduction of new varieties with high resilience to stress. This covers resistance to pests and diseases, as well as to stress caused by conditions such as heat and drought. The emphasis is on developing new measurement and analysis methods using sensors, crop models and artificial intelligence. Organisations from nine countries are participating in the project.The Netherlands is represented by Naktuinbouw, NAK and WUR. The focus in the Netherlands is on the Workpackage 2 sub-project: setting up a fast, cheap method to automate the measurement procedure of a number of external traits.

 

More about INVITE can be found here: INVITE: Innovations in variety testing