Scholarlink

<p>Traditional tomato ripeness detection mostly relies on single-color or single-spectral technologies, which have limitations such as limited detection accuracy and susceptibility to environmental interference. Since tomato ripeness significantly affects its quality indicators, ripeness determination is crucial in the agricultural production and food processing industries. In this study, tomato spectral data and ripeness characterization data were collected, and Gaussian Process Regression (GPR), Partial Least Squares Regression (PLSR), and Support Vector Regression (SVR) were used to quantitatively characterize the spectral features associated with tomato ripeness. The optimal combination of preprocessing methods, feature wavelength extraction methods and modeling methods was selected through comparative experiments. Based on this optimal modeling scheme, hyperspectral feature wavelengths were fused with the hue (H) values in the HSV color space, and the detection performance of the One-Dimensional Convolutional Neural Network (1D-CNN) and machine learning methods was compared. In addition, the Convolutional Block Attention Module (CBAM) was incorporated to construct the 1D-CBAMCNN model for tomato ripeness detection, which achieved an accuracy of 95.18%, a 2-percentage-point improvement compared with the benchmark model 1D-CNN. Meanwhile, the SHapley Additive exPlanations (SHAP) method was adopted to analyze the importance of feature wavelengths, enhancing the interpretability of the model. This study provides a feasible solution for high-precision and interpretable non-destructive detection of tomato ripeness, and has practical significance for promoting the development of precision agriculture and post-harvest processing technologies.</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

Tomato ripeness detection using hyperspectral imaging technology with multimodal fusion and improved 1D-CNN models

  • 期刊论文

  • 发表日期：

    2026年3月13日

查看全文
  • Qin Zhou,
  • Shiai Zhou,
  • Xiaohuan Li,
  • Yiding Zhang,
  • Lingxian Zhang

摘要

Traditional tomato ripeness detection mostly relies on single-color or single-spectral technologies, which have limitations such as limited detection accuracy and susceptibility to environmental interference. Since tomato ripeness significantly affects its quality indicators, ripeness determination is crucial in the agricultural production and food processing industries. In this study, tomato spectral data and ripeness characterization data were collected, and Gaussian Process Regression (GPR), Partial Least Squares Regression (PLSR), and Support Vector Regression (SVR) were used to quantitatively characterize the spectral features associated with tomato ripeness. The optimal combination of preprocessing methods, feature wavelength extraction methods and modeling methods was selected through comparative experiments. Based on this optimal modeling scheme, hyperspectral feature wavelengths were fused with the hue (H) values in the HSV color space, and the detection performance of the One-Dimensional Convolutional Neural Network (1D-CNN) and machine learning methods was compared. In addition, the Convolutional Block Attention Module (CBAM) was incorporated to construct the 1D-CBAMCNN model for tomato ripeness detection, which achieved an accuracy of 95.18%, a 2-percentage-point improvement compared with the benchmark model 1D-CNN. Meanwhile, the SHapley Additive exPlanations (SHAP) method was adopted to analyze the importance of feature wavelengths, enhancing the interpretability of the model. This study provides a feasible solution for high-precision and interpretable non-destructive detection of tomato ripeness, and has practical significance for promoting the development of precision agriculture and post-harvest processing technologies.