LeTra: Tracking Plant Growth with AI and Smart Overlap Measures

Jim Crocker
18th January, 2024

LeTra: Tracking Plant Growth with AI and Smart Overlap Measures

Image Source: Natural Science News, 2024

Understanding how plants turn sunlight into energy through photosynthesis is crucial for improving crop yields and feeding our growing population. This process has traditionally been hard to study on a large scale, but technology is changing that. Researchers can now employ automated high-throughput phenotyping (HTP) systems to measure photosynthesis more consistently and efficiently. These systems use chlorophyll fluorescence imaging to assess plants but often rely on humans to interpret leaf-level data. The demand for fully automated leaf tracking is therefore on the rise. Existing methods for tracking the growth and behavior of individual leaves over time are either too complex, require massive amounts of data, or are just emerging. Accessible applications and assorted techniques are necessary to move forward. With more advanced phenotyping platforms being developed, integrating sophisticated computer vision technologies like convolutional neural networks is vital but slow-going. Now, researchers have developed a method designed to improve leaf tracking using top-down images of plants. This new method relies on fine-tuning a type of neural network known as Mask R-CNN, which can distinguish and track objects in images. By improving how these networks interpret the overlap of leaves in images, through what's called intersection over union, plants can be effectively monitored. The researchers have also shared datasets and codes needed to further refine this technology. The goal is to encourage others in the field to develop and enhance current leaf-tracking methodologies. When tested for their accuracy in identifying and outlining leaves, this method scored impressively high—for detection, the mean score was 0.956 and for segmentation overlap, it was 0.844. Tracking capabilities were also robust, showing a success rate of approximately 84% and a higher order tracking accuracy rating of 0.846 across various plants and leaves. In a practical example, the researchers found out that the age and order of leaves on a plant could affect its photosynthetic ability and how it responds to different light conditions. It turns out that leaves might perform photosynthesis differently depending on the plant's genetic makeup. The strength of this tracking method is that it requires relatively few data samples to start providing good results, thanks to the preliminary adjustments made to the neural network. Most of the tracking problems encountered could potentially be solved simply by providing more data to train the Mask R-CNN model. In summary, this new technology developed by the Center for Research in Agricultural Genomics offers a powerful tool for automatic leaf tracking, which can lead to more nuanced insights into plant photosynthesis and ultimately, better crop management strategies.

AgricultureBiotechBiochem

References

Main Study

1) LeTra: a leaf tracking workflow based on convolutional neural networks and intersection over union.

Published 17th January, 2024

https://doi.org/10.1186/s13007-024-01138-x


Related Articles

An unhandled error has occurred. Reload 🗙