Corn seed dataset based on hyperspectral and RGB images.

Abstract

This study employed an HY-6010-S hyperspectral imaging system, covering a spectral range of 400-1000 nm, combined with an RGB industrial camera to acquire multimodal data. The dataset simulates phenotypic analysis scenarios of maize seeds under controlled laboratory conditions, with the ambient temperature maintained at 20-25°C. Comprehensive testing was conducted using 12 different maize varieties. Approximately 200 seed samples were collected per variety, resulting in a total sample size of about 2400, each subjected to hyperspectral and RGB image acquisition. Preprocessing steps included noise reduction, background removal, band selection, and modality alignment. To ensure the accuracy and reliability of the experimental data, HHIT software and Python were utilized for data processing. This dataset plays a significant role in seed variety classification, phenotypic analysis, precision agriculture, and machine learning applications.

Code and data availability

The paper is a Data in Brief article depositing its own multimodal maize seed hyperspectral and RGB image dataset (2400 seeds, 12 varieties) on Mendeley Data, with a direct public URL and DOI given in the article.