Assessing sensors for enhanced canola plant architecture selection

New technology is being explored for its application in determining yield components for canola to accelerate yield gains.

Yield potential in water-limited environments depends on crop transpiration, transpiration efficiency and harvest index. Canola has a lower harvest index compared to cereal crops, which is especially pronounced in water-limited conditions.

Improving yield components in canola has been challenging due to difficulties in selecting promising genetic material using standard techniques.

Next-generation phenomic technologies could help overcome these challenges by developing high-throughput methods for selecting traits that optimise harvest index and yield in canola.

One method is to estimate biomass in the paddock non-destructively using combined sensor data. To this end, GRDC is supporting a pilot study with a Dutch company, PhenoKey, which designs and implements tailored automation solutions. The pilot study is being conducted in collaboration with Fraunhofer Institute for Integrated Circuits - Embedded Systems and Real-Time Computing, , a division of the Fraunhofer Institute IIS, which specialises in the development and application of innovative X-ray and computed tomography (CT) technologies. In this instance, PhenoKey and Fraunhofer are undertaking a feasibility study using a phenotyping platform to collect data from an X-ray system, light detection and ranging (LiDAR) and GPS sensor.

Figure 1: GRDC is undertaking a feasibility study with a Dutch company, PhenoKey, and the German-based Fraunhofer IIS to evaluate a phenotyping platform with different sensors: X-ray system, LiDAR and RTK-GPS sensor for use in canola.

Photo: Fraunhofer IIS

The goal is to assess the biomass of yield components on a paddock scale for different canola varieties using this technology and compare the sensor-derived data with the actual measured biomass.

If the correlation between the two biomass values is robust, this project would be the basis of developing a rapid phenotyping method to assess harvest index in canola. Such a high-throughput method of phenotyping would provide a useful tool to accelerate the breeding of higher-yielding canola varieties.

Methods

The phenotyping platform used in this study is a remotely controlled field vehicle equipped with three different sensors. The sensors are an X-ray source with an X-ray detector, a LiDAR sensor and an RTK-GPS sensor (refer main photo).

The X-ray sensor measures the absorbance of the plants between the source and detector in 2D projections as the vehicle is driving through the paddock. The amount of absorbed X-ray intensity is for plant material highly correlated with the physical density of the plants.

The measured density values are used to estimate the virtual biomass, which later will be correlated with the actual biomass. As the vehicle does not have constant velocity, due for instance to ground conditions, the RTK-GPS data is used to stretch or compress the 2D X-ray images (Figure 2).

Figure 2: X-ray data of a canola plot stretched with RTK-GPS data over three metres.

Credit: Fraunhofer IIS

Furthermore, the GPS data is used to link the X-ray data to the different plots to obtain a biomass value per plot, which can later be compared to the actual measured biomass after harvesting. The LiDAR sensor generates a point cloud of the plot in a 3D-coordinate system. This data is used to estimate the total volume of each plot from the soil to the top. The volume is used to extrapolate the value of the estimated biomass derived from the X-ray data, since the X-ray sensor cannot scan the full volume of the plot.

Initial findings and next steps

The results from studies in 2023 showed that the sensor-derived data is suitable for the non-destructive estimation of virtual total biomass per plot in a canola paddock.

In 2024 a more in-depth appraisal of canola biomass production will be undertaken in this feasibility study. Using 36 plots containing six varieties, four scanning campaigns are planned: before flowering, after flowering, at the beginning of pod formation and shortly before harvest.

This will enhance the dataset and strengthen the assessment of this phenotyping method, validating its correlation with actual measured biomass in canola, a step forward in creating high-throughput, breeding program-friendly methods for selecting traits that optimise harvest index and yield in canola.