3D reconstruction and morphological characterisation of single wheat grains by X-ray μCT

Abstract

Abstract The intricate task of achieving three-dimensional (3D) visual reconstruction of wheat kernels represents a notable challenge within the domain of digital grain analysis, playing a pivotal role in the realms of grain storage, processing, and breeding. However, existing investigations focused on individual kernels predominantly encompass dimensions such as length, width, height, and epidermal texture features, with a tendency to be invasive to the internal organisational structure of the kernel. Non-local mean filtering algorithm is proposed to segment various tissues, and the 2D grey scale images are extracted from X-ray micro-computed tomography (μCT) to reconstruct a meticulous 3D visualisation model of individual wheat grains. Furthermore, building upon this foundation, an exhaustive assessment of morphological and structural parameters pertaining to each facet of the internal organisation of the wheat seed grain is conducted. Notably, these calculated parameters align with data generated by prior researchers，with 80% of the volume of the endosperm, 12% of the pericarp, about 2% of the endosperm and scutellum, and 4% of the pores. The established parameters and resultant 3D visual models serve as foundational components for subsequent in-depth examinations into various physicochemical properties, including quality characteristics, heat, and mass transfer attributes, as well as variations in its morphological structure during breeding, which are pertinent to individual grains. This research contributes valuable insights and methodologies that can propel the advancement of studies in wheat kernel analysis.

Plant phenotyping relevance

X線μCT画像から小麦粒の3D形状・内部組織を再構成し、形態・構造パラメータを抽出する手法が研究の中心であり、植物器官形質の取得方法を開発・評価している。

Code and data availability

The paper's CT image dataset (>27,000 slices) and 3D models are not publicly deposited; the data availability statement says they are available only on request from the corresponding author. No author code or public repository URL is provided.

No evidence-backed public reproduction asset is currently recorded.