News — This approach significantly improves the accuracy of leaf morphology analysis, providing new opportunities for understanding plant growth and optimizing agricultural productivity.
Leaves, essential for photosynthesis and other physiological functions, exhibit diverse shapes that help plants adapt to their environments. Traditional methods for measuring leaf morphology have relied heavily on 2D imaging, which fails to capture the complex 3D structures crucial for understanding leaf functionality. These limitations have driven the development of more sophisticated 3D imaging techniques, yet challenges persist in accurately mapping leaf edges, particularly in highly detailed structures.
A published in on 9 May 2024, provides a new way to analyze leaf morphology nondestructively, enabling precise mapping of plant structures in 3D.
The research utilized a novel method for 3D leaf edge reconstruction, combining deep-learning-based 2D instance segmentation with curve-based 3D reconstruction techniques. The approach was first tested on virtual leaf models under ideal conditions, reconstructing 3D leaf edges by extracting 2D edges from true mask images. Performance was influenced by the support threshold (τt), where low values led to inaccuracies and high values caused incomplete coverage. For scenes with multiple leaves, leaf correspondence identification was critical to ensure accurate reconstructions despite occlusions. Accuracy assessments showed that larger leaves were easier to reconstruct, while small leaves with greater curvature posed difficulties. The method was sensitive to camera noise but less so to the number of images or occlusion levels, suggesting that additional images did not necessarily enhance precision. Using Mask R-CNN for real plant data, individual leaf masks were generated, though smaller leaves presented challenges, reflected by lower average precision (AP) scores. When applied to actual soybean plants at various growth stages, the method successfully reconstructed most leaves at optimal support thresholds but faced difficulties with highly occluded or small leaves, sometimes resulting in artifacts or shape distortions. Testing on diverse leaf types revealed that while the method accurately captured lobed edges, serrated and elongated leaves exhibited reduced detail, especially at the apex. Leaves with up to three holes were reconstructed well, but precision declined with more holes. Overall, the approach effectively captured complex leaf morphologies but required further refinement for intricate or highly occluded structures.
According to the study's lead researcher, Dr. Koji Noshita,“Our method represents a breakthrough in 3D leaf edge reconstruction. By combining deep learning and curve-based techniques, we can now achieve a much higher level of detail in leaf morphology analysis, which has important implications for agricultural research.”
This 3D leaf edge reconstruction technique marks a significant advancement in plant phenotyping. It opens up new avenues for studying plant morphology in greater detail, with broad applications in agriculture and ecological research. The researchers believe that continued development of this method will pave the way for more efficient and productive agricultural practices, addressing global food security challenges.
###
References
DOI
Original Source URL
Funding information
This study was supported by Japan Society for the Promotion of Science (JSPS) KAKENHI Grant Numbers 20H01381, 21K14947, and 22H04727 (to K.N.); Japan Science and Technology Agency (JST) PRESTO Grant Number JPMJPR16O5 (to K.N.); JST MIRAI Grant Number JPMJMI20G6 (to K.N.); Moonshot R&D Grant Number JPMJMS2021 (to K.N.); and Bio-oriented technology Research Advancement InstitusioN (BRAIN) Moonshot R&D Grant Number JPJ009237 (to K.N.).
About
Science Partner Journal is an online-only Open Access journal published in affiliation with the State Key Laboratory of Crop Genetics & Germplasm Enhancement, Nanjing Agricultural University (NAU) and distributed by the American Association for the Advancement of Science (AAAS). Like all partners participating in the Science Partner Journal program, Plant Phenomics is editorially independent from the Science family of journals. Editorial decisions and scientific activities pursued by the journal's Editorial Board are made independently, based on scientific merit and adhering to the highest standards for accurate and ethical promotion of science. These decisions and activities are in no way influenced by the financial support of NAU, NAU administration, or any other institutions and sponsors. The Editorial Board is solely responsible for all content published in the journal. To learn more about the Science Partner Journal program, visit the SPJ program homepage.
The mission of is to publish novel research that will advance all aspects of plant phenotyping from the cell to the plant population levels using innovative combinations of sensor systems and data analytics. Plant Phenomics aims also to connect phenomics to other science domains, such as genomics, genetics, physiology, molecular biology, bioinformatics, statistics, mathematics, and computer sciences. Plant Phenomics should thus contribute to advance plant sciences and agriculture/forestry/horticulture by addressing key scientific challenges in the area of plant phenomics.
RSS