Scholarlink

<p>Plant phenotyping is the study of the agronomics traits of plants, advanced into a data-driven science that integrates imaging and machine learning. These approaches are increasingly applied in plant phenotyping for crop monitoring and water status detection, thus optimizing water use and improving breeding strategies. The potential of smartphone RGB imaging in discriminating maize plants under irrigation regimes and revealing image-related features with respect to water status and chlorophyll was evaluated. Indices were computed from R-G-B values: those derived from RGB include numerous vegetation indices: Hue, Saturation, Intensity, and Dark Green Color Index (DGCI). Results show that different chlorophyll contents were strongly linearly correlated with indices of DGCI (<i>r</i> = 0.81, <i>p</i> &lt; 0.01), Hue (<i>r</i> = -0.76, <i>p</i> &lt; 0.01), and Saturation (<i>r</i> = 0.72, <i>p</i> &lt; 0.05), confirming their sensitivity to drought stress. various multivariate classification models easily differentiated treatments with PCA, explaining 68.4% of the total variance in the first two components. PLS-DA achieved classification accuracy of 85.7%, while Random Forest successfully distinguished between treatments with the highest performance (overall accuracy = 92.4%, kappa = 0.89). The feature importance analysis reveals that DGCI, ExG, and LB are the most informative features for water status. This study highlights smartphone-based phenotyping as a reliable technique for nondestructive monitoring at a low cost, thus having robust applications for precision irrigation and sustainable crop management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Image segmentation and machine learning modelling for water status determination and precision irrigation management

  • 期刊论文

  • 发表日期：

    2025年10月24日

查看全文
  • Hamza Salah Ud Din,
  • Aqib Iqbal,
  • Fazal Jalal,
  • Awais Ahmad,
  • Muhammad Haris Paracha,
  • Shah Fahad,
  • Zafar Hayat Khan,
  • Shah Saud,
  • Taufiq Nawaz

摘要

Plant phenotyping is the study of the agronomics traits of plants, advanced into a data-driven science that integrates imaging and machine learning. These approaches are increasingly applied in plant phenotyping for crop monitoring and water status detection, thus optimizing water use and improving breeding strategies. The potential of smartphone RGB imaging in discriminating maize plants under irrigation regimes and revealing image-related features with respect to water status and chlorophyll was evaluated. Indices were computed from R-G-B values: those derived from RGB include numerous vegetation indices: Hue, Saturation, Intensity, and Dark Green Color Index (DGCI). Results show that different chlorophyll contents were strongly linearly correlated with indices of DGCI (r = 0.81, p < 0.01), Hue (r = -0.76, p < 0.01), and Saturation (r = 0.72, p < 0.05), confirming their sensitivity to drought stress. various multivariate classification models easily differentiated treatments with PCA, explaining 68.4% of the total variance in the first two components. PLS-DA achieved classification accuracy of 85.7%, while Random Forest successfully distinguished between treatments with the highest performance (overall accuracy = 92.4%, kappa = 0.89). The feature importance analysis reveals that DGCI, ExG, and LB are the most informative features for water status. This study highlights smartphone-based phenotyping as a reliable technique for nondestructive monitoring at a low cost, thus having robust applications for precision irrigation and sustainable crop management.