[Paper Review] Image-based phenotyping of diverse Rice (Oryza Sativa L.) Genotypes

Mukesh Kumar Vishal, Dipesh Tamboli+9 more|arXiv (Cornell University)|Apr 6, 2020

Smart Agriculture and AI14 references4 citations

TL;DR

This study proposes an image-based phenotyping framework using YOLO-based deep learning for high-throughput leaf counting and morphological trait extraction in 10 diverse rice (Oryza sativa L.) genotypes. The method enables accurate discrimination of drought-tolerant and susceptible genotypes through traits like leaf count, convex hull area, and canopy spread, achieving robust clustering via Ward’s method with high phenotypic resolution under controlled conditions.

ABSTRACT

Development of either drought-resistant or drought-tolerant varieties in rice (Oryza sativa L.), especially for high yield in the context of climate change, is a crucial task across the world. The need for high yielding rice varieties is a prime concern for developing nations like India, China, and other Asian-African countries where rice is a primary staple food. The present investigation is carried out for discriminating drought tolerant, and susceptible genotypes. A total of 150 genotypes were grown under controlled conditions to evaluate at High Throughput Plant Phenomics facility, Nanaji Deshmukh Plant Phenomics Centre, Indian Council of Agricultural Research-Indian Agricultural Research Institute, New Delhi. A subset of 10 genotypes is taken out of 150 for the current investigation. To discriminate against the genotypes, we considered features such as the number of leaves per plant, the convex hull and convex hull area of a plant-convex hull formed by joining the tips of the leaves, the number of leaves per unit convex hull of a plant, canopy spread - vertical spread, and horizontal spread of a plant. We trained You Only Look Once (YOLO) deep learning algorithm for leaves tips detection and to estimate the number of leaves in a rice plant. With this proposed framework, we screened the genotypes based on selected traits. These genotypes were further grouped among different groupings of drought-tolerant and drought susceptible genotypes using the Ward method of clustering.

Motivation & Objective

• To develop a high-throughput, non-destructive phenotyping pipeline for discriminating drought-tolerant and drought-susceptible rice genotypes.
• To address limitations of traditional phenotyping, such as subjectivity, time consumption, and low throughput, by leveraging automated image analysis.
• To quantify key morphological traits—leaf count, convex hull area, canopy spread, and leaf-to-convex-hull ratio—for diverse rice genotypes.
• To evaluate the performance of YOLOv3 for accurate leaf tip detection and leaf counting in rice plants under controlled greenhouse conditions.
• To group genotypes into drought-tolerant and susceptible clusters using hierarchical clustering (Ward’s method) based on image-derived phenotypic traits.

Proposed method

• Collected top-view RGB images of 10 rice genotypes at 5-day intervals using a LemnaTec scanalyzer3D system at NDPPC, IARI, New Delhi.
• Pre-processed and resized images (6576×4384) and manually annotated leaf tips on ~300 images for YOLOv3 training.
• Trained YOLOv3 (You Only Look Once) for real-time object detection of leaf tips, enabling fast and accurate leaf counting.
• Calculated morphological traits: number of leaves per plant, convex hull area, canopy spread (horizontal and vertical), and leaf-to-convex-hull ratio.
• Applied Ward’s linkage hierarchical clustering to group genotypes based on phenotypic similarity using the extracted image-based traits.
• Used a subset of 150 genotypes from a GWAS study, focusing on high polymorphism in leaf number, under moisture-deficit (12.5%) and well-watered (25%) conditions.

Experimental results

Research questions

• RQ1 Can YOLO-based deep learning accurately detect leaf tips and estimate leaf count in rice plants from top-view RGB images?
• RQ2 Which image-derived morphological traits (e.g., convex hull area, canopy spread, leaf-to-convex-hull ratio) best differentiate drought-tolerant from drought-susceptible rice genotypes?
• RQ3 To what extent can image-based phenotyping improve the precision and throughput of drought tolerance screening compared to traditional methods?
• RQ4 How effective is Ward’s hierarchical clustering in grouping rice genotypes based on image-derived phenotypic traits under controlled stress conditions?
• RQ5 Can the ratio of leaves per unit convex hull area serve as a reliable indicator of plant growth efficiency and stress response in early vegetative stages?

Key findings

• YOLOv3 achieved high accuracy in leaf tip detection and leaf counting, enabling reliable phenotypic quantification in rice plants under controlled conditions.
• The number of leaves per unit convex hull area emerged as a key discriminative trait, showing significant variation between drought-tolerant and susceptible genotypes.
• Canopy spread (horizontal and vertical) measurements showed consistent differences across genotypes, supporting their use in phenotypic clustering.
• Ward’s method successfully grouped the 10 rice genotypes into distinct clusters of drought-tolerant and drought-susceptible types based on image-derived traits.
• The proposed framework demonstrated high throughput and reproducibility, reducing human error and time compared to conventional phenotyping.
• The integration of deep learning with morphological trait analysis enabled non-invasive, high-resolution phenotyping suitable for early-stage stress response monitoring in rice.