PLS-DA model for accurate identification of Chinese cabbage leaf color based on multispectral imaging

Abstract

Chinese cabbage (Brassica rapa L. ssp. pekinensis), a leafy vegetable, exhibits a range of leaf colors, with the dark green varieties being favored by consumers. Manual visual identification of Chinese cabbage leaf color phenotypes is subjective and it is difficult to distinguish between subtle differences in leaf color, posing challenges for precision breeding. In this study, we constructed a partial least squares discriminant analysis (PLS-DA) leaf color identification model and compared four classification methods for leaf color, namely red, green, and blue (RGB) channels, hue, saturation, and lightness (HSL) color space, multi-spectrum and data-fusion. The PLS-DA supervised leaf color phenotype identification model based on data fusion can improve the recognition rate by 1%−13% compared to a single spectral model. To further validate the model, we conducted a bulked segregant analysis (BSA) of a mixed pool of a Chinese cabbage F2 population (F2-449) using whole-genome sequencing. The candidate locus related to dark green leaf color was reduced by 9.76 Mb compared to the manual visual inspection which provides convenience for the localization of candidate genes. Therefore, the development of a precise phenotypic identification system for Chinese cabbage that can distinguish subtle leaf color differences using high-throughput phenotype analysis technology is of great significance and agricultural practical value for the mining of high-throughput genomic data.

Plant phenotyping relevance

マルチスペクトル画像とPLS-DAを用いて、白菜の葉色という植物表現型を高スループットかつ高精度に識別する手法を開発・検証しており、表現型取得・抽出法が研究の中心である。

Code and data availability

The paper describes a PLS-DA leaf-color identification model built from VideometerLab 4 multispectral imaging of Chinese cabbage leaves, but no supplied block contains a public phenotype dataset, image/spectral data deposit, author analysis code, or trained model with an availability statement or public URL. Supplement

No evidence-backed public reproduction asset is currently recorded.