A Salt-Tolerance evaluation system for Chinese cabbage using multispectral image data fusion and Fine-Tuned, pruned convolutional-LSTM-ResNet networks

Abstract

Global soil salinization presents an increasing threat to vegetable productivity and agricultural yields. The Chinese cabbage (Brassica rapa subsp. pekinensis) is a vital vegetable crop in China and across many regions in Asia. However, its quality and yield are highly susceptible to salt stress. Consequently, developing an efficient and accurate evaluation system for screening salt-tolerant Chinese cabbage varieties is essential. This study proposes a dual-input data fusion model for salt tolerance evaluation. The first input consists of one-dimensional sequence data derived from the region of interest texture features and spectral data, processed using convolutional neural network (CNN) and long short-term memory. The second input comprises two-dimensional multispectral images, analyzed through a CNN and ResNet network, optimized using fine-tuning and pruning techniques. These inputs were independently processed in a dual-branch network, with their outputs fused in a fully connected layer to deliver a comprehensive assessment of salt tolerance. A comparative analysis with a photosynthetic phenotype imaging system revealed the superior information richness and accuracy of the proposed model. Validation of the salt tolerance classification achieved an accuracy of 95.00% on the Day 5 following salt stress, with only four varieties misclassified, underscoring the efficiency and effectiveness of the model in early screening. On Day 9 following salt stress, all salt-sensitive varieties were fully identified. Using this model, we identified seven salt-tolerant, seven salt-neutral, and five salt-sensitive Chinese cabbage varieties. Integrating one-dimensional and two-dimensional data enabled the extraction of key plant parameters, such as chlorophyll content, growth status, and leaf structure. This approach provides an effective tool for evaluating salt tolerance in Chinese cabbage and other vegetables, facilitating breeding efforts and mitigating the effects of soil salinization on global crop yield.

Plant phenotyping relevance

植物のマルチスペクトル画像と画像由来特徴を用いて塩耐性を評価するデータ融合・深層学習システムを開発し、既存の表現型イメージングシステムと比較検証しているため、表現型取得・抽出法が中心である。

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.