Hyperspectral image processing and analysis notebook for wheat grain characterisation
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Purpose and Scope: This Jupyter notebook implements a comprehensive processing pipeline for hyperspectral image (HSI) analysis acquired using short-wave infrared (SWIR) imaging technology. Specifically designed for the Watkins wheat project, the pipeline enables automated characterisation and spectral profiling of individual wheat grains.Key Processing Steps:Hyperspectral data import and visualisation (ENVI format)Bad pixel detection and removal via median filteringRadiometric calibration using dark and white reference measurementsData normalisation and absorbance conversionAutomated grain segmentation using Otsu thresholdingIndividual grain labeling and spectral feature extractionBatch spectral profile analysis and exportTechnical Implementation: The notebook leverages Python scientific libraries including spectral, scikit-image, NumPy, SciPy, and Pandas for processing multi-dimensional hyperspectral datasets. Morphological operations and connected-component labeling enable robust grain detection and isolation from background.Data Output: Generates calibrated absorbance data, binary segmentation masks, individual grain spectral profiles, and CSV exports of spectral signatures across all spectral bands (wavelengths).Institutional Context: Developed collaboratively by the University of Adelaide and University of Nottingham for wheat breeding and phenotyping research. BB/W006979/1, BB/W005751/1, BB/W005972/1, BB/V018108/1, BB/X018806/1] and the Australian Research Council [grant number FT210100810]
Fields of Study & Subjects
Spectroscopy and Chemometric Analyses
Smart Agriculture and AI
Remote-Sensing Image Classification
Publication details
- Journal / source
- Figshare
- Publication date
- 1 January 2026
- Language
- EN