Scholarlink

<p>Potential use of novel rapid and non-destructive Vis–NIR spectroscopy with multispectral imaging (MSI) system between 405 and 970&#xa0;nm has gained the interest of chemical composition prediction of under processed foods. This study aimed to develop more advanced models coupled with spectra (405–970&#xa0;nm) to enhance prediction stability and easy handling of non-linear complex spectra. Models were developed such as genetic algorithm (GA) and whale optimization algorithm (WOA) algorithms of partial least square (PLS) and back propagation neural netWOArk (BPNN), support vector machine (SVM), and feedforward neural network (FNN) to handle more complex non-linear data. Total water prediction was conducted in mushroom under different dehydration methods as hot-air (HD), far-infrared (FIR), and freeze-drying (FD) at a constant temperature of 70&#xa0;°C. This study demonstrated HD rapidly reduced water with final contents of 8.68% comparing with FIR and FD of after 240&#xa0;min. Developed models performed well for FD followed by FIR due to high water contents. Among all seven models, FNN prediction efficiency was the&#xa0;highest and obtained coefficient of determination (<i>R</i><sup>2</sup><sub>p</sub>) of 0.9794 with lowest root mean square error for prediction (RMSEP) from 4.6781. In terms of model robustness, GABPNN achieved higher RPD values between 5.01 and 9.54. VIS–NIR spectroscopy through MSI is a promising tool combined with chemometrics for online assessment of food quality attributes.</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

Comparison of FNN with Advanced Algorithms in Non-destructive Estimation of Water State of Mushroom During Thermal Processing

  • 期刊论文

  • 发表日期：

    2025年11月19日

查看全文
  • Shoaib Younas,
  • Farhan Ali,
  • Ukasha Arqam,
  • Xin Wang,
  • Fatima Tariq,
  • Asif Wali

摘要

Potential use of novel rapid and non-destructive Vis–NIR spectroscopy with multispectral imaging (MSI) system between 405 and 970 nm has gained the interest of chemical composition prediction of under processed foods. This study aimed to develop more advanced models coupled with spectra (405–970 nm) to enhance prediction stability and easy handling of non-linear complex spectra. Models were developed such as genetic algorithm (GA) and whale optimization algorithm (WOA) algorithms of partial least square (PLS) and back propagation neural netWOArk (BPNN), support vector machine (SVM), and feedforward neural network (FNN) to handle more complex non-linear data. Total water prediction was conducted in mushroom under different dehydration methods as hot-air (HD), far-infrared (FIR), and freeze-drying (FD) at a constant temperature of 70 °C. This study demonstrated HD rapidly reduced water with final contents of 8.68% comparing with FIR and FD of after 240 min. Developed models performed well for FD followed by FIR due to high water contents. Among all seven models, FNN prediction efficiency was the highest and obtained coefficient of determination (R2p) of 0.9794 with lowest root mean square error for prediction (RMSEP) from 4.6781. In terms of model robustness, GABPNN achieved higher RPD values between 5.01 and 9.54. VIS–NIR spectroscopy through MSI is a promising tool combined with chemometrics for online assessment of food quality attributes.