Hyperspectal imaging technology for phenotyping iron and boron deficiency in Brassica napus under greenhouse conditions.

Abstract

Introduction The micronutrient deficiency of iron and boron is a common issue affecting the growth of rapeseed ( Brassica napus ). In this study, a non-destructive diagnosis method for iron and boron deficiency in Brassica napus (genotype: Zhongshuang 11) using hyperspectral imaging technology was established. Methods The recognition accuracy was compared using the Fisher Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) recognition models. Recognition results showed that Multiple Scattering Correction (MSC) could be applied for the full band hyperspectral data processing, while the LDA models presented better performance on establishing the leaf iron and boron deficiency symptom recognition than the SVM models. Results The recognition accuracy of the training set reached 96.67%, and the recognition rate of the prediction set could be 91.67%. To improve the model accuracy, the Competitive Adaptive Reweighted Sampling algorithm (CARS) was added to construct the MSC-CARS-LDA model. 33 featured wavelengths were selected via CARS. The recognition accuracy of the MSC-CARS-LDA training set was 100%, while the recognition accuracy of the MSC-CARS-LDA prediction set was 95.00%. Discussion This study indicates that, it is capable to identify the iron and boron deficiency in rapeseed using hyperspectral imaging technology.

Plant phenotyping relevance

アブラナの鉄・ホウ素欠乏症状をハイパースペクトル画像から非破壊認識する手法を開発し、LDA/SVMや特徴波長選択で精度検証しており、表現型取得・抽出が研究の中心である。

Code and data availability

The supplied blocks describe hyperspectral imaging of Brassica napus under iron/boron deficiency with LDA/SVM modeling, but contain no public dataset, image, code, or model deposit. No data availability statement text with an authors' public URL appears in the supplied blocks; the only external URL is the USDA Rapeseed

No evidence-backed public reproduction asset is currently recorded.