Scholarlink

<p>As a key indicator for evaluating the nutritional value and quality of fruit, the precise characterization of the spatial distribution of fruit polyphenols is essential for regulating fruit cultivation, postharvest grading, and processing optimization. Conventional methods for polyphenol content detection suffer from destructiveness, low efficiency, and an inability to visualize spatial distribution. Although existing detection models based on spectral analysis and machine learning primarily focus on overall content prediction, they struggle to accurately capture micro-regional distribution differences in fruit polyphenols. To address these issues, this study proposes the hyperspectral imaging-one-dimensional convolutional neural network-fruit polyphenol distribution visualization (HSI-1D CNN-FPDV) model. The model acquires global spectral data of fruits via the HSI system and optimizes the spectral preprocessing procedure using an adaptive noise suppression algorithm. A multi-scale one-dimensional convolutional neural network architecture is constructed to enhance the mapping between spectral features and polyphenol content. Furthermore, a collaborative mechanism of pixel-level content inversion and visualization rendering is designed to accurately map polyphenol distribution. Simulation and experimental results show that the prediction accuracy of polyphenol content reaches 98.7%, which is 16.2% and 13.5% higher than that of the comparative methods, respectively. The spatial resolution achieves 0.12&#xa0;mm, representing a 45.2% improvement over traditional spectral models. Moreover, the visualization time for a single fruit is reduced to 6.8&#xa0;s, which is 58.3% and 52.1% lower than that of the comparative methods, respectively. The model effectively adapts to apples, grapes, and cherries, improving versatility by 58%. This approach overcomes the destructive and low-precision limitations inherent in conventional polyphenol distribution detection methods and provides an alternative framework for non-destructive, accurate fruit quality evaluation.</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

Visualization model of fruit polyphenol content distribution based on hyperspectral imaging and one-dimensional convolutional neural network

  • 期刊论文

  • 发表日期：

    2026年7月1日

查看全文
  • Shenyan Li

摘要

As a key indicator for evaluating the nutritional value and quality of fruit, the precise characterization of the spatial distribution of fruit polyphenols is essential for regulating fruit cultivation, postharvest grading, and processing optimization. Conventional methods for polyphenol content detection suffer from destructiveness, low efficiency, and an inability to visualize spatial distribution. Although existing detection models based on spectral analysis and machine learning primarily focus on overall content prediction, they struggle to accurately capture micro-regional distribution differences in fruit polyphenols. To address these issues, this study proposes the hyperspectral imaging-one-dimensional convolutional neural network-fruit polyphenol distribution visualization (HSI-1D CNN-FPDV) model. The model acquires global spectral data of fruits via the HSI system and optimizes the spectral preprocessing procedure using an adaptive noise suppression algorithm. A multi-scale one-dimensional convolutional neural network architecture is constructed to enhance the mapping between spectral features and polyphenol content. Furthermore, a collaborative mechanism of pixel-level content inversion and visualization rendering is designed to accurately map polyphenol distribution. Simulation and experimental results show that the prediction accuracy of polyphenol content reaches 98.7%, which is 16.2% and 13.5% higher than that of the comparative methods, respectively. The spatial resolution achieves 0.12 mm, representing a 45.2% improvement over traditional spectral models. Moreover, the visualization time for a single fruit is reduced to 6.8 s, which is 58.3% and 52.1% lower than that of the comparative methods, respectively. The model effectively adapts to apples, grapes, and cherries, improving versatility by 58%. This approach overcomes the destructive and low-precision limitations inherent in conventional polyphenol distribution detection methods and provides an alternative framework for non-destructive, accurate fruit quality evaluation.