Scholarlink

<p>Traditional lamb quality testing relies on destructive physical and chemical analyses that are time-consuming and require strict experimental conditions. Hyperspectral imaging offers a non-destructive, rapid, and high-throughput alternative but faces issues such as data anomalies and weak predictions of nonlinear quality traits. Using chilled lamb, we established a spectral model for physical quality characterization based on hardness (g) and elasticity (g/sec). A second-iteration Monte Carlo sampling method was designed to improve data quality and reduce abnormal sample misclassification. We employed the statistical approach to filter irrelevant and redundant information from spectral data by using coupled t-detection and contribution measures. On this basis, a dimensionality reduction partial least squares regression (DR-PLSR) model was developed for quantitative prediction of lamb physical properties. The optimal model achieved an <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\:{R}_{P}^{2}\)</EquationSource> </InlineEquation> of 0.986, an <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\:\text{R}\text{M}\text{S}\text{E}\text{P}\)</EquationSource> </InlineEquation> of 23.986, and an RPD of 7.483 for hardness prediction, while the best elasticity model yielded an <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(\:{R}_{P}^{2}\)</EquationSource> </InlineEquation> of 0.812, an <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(\:\text{R}\text{M}\text{S}\text{E}\text{P}\)</EquationSource> </InlineEquation> of 0.016, and an RPD of 3.022. These results demonstrate the effectiveness of hyperspectral imaging combined with DR-PLSR for non-destructive assessment of chilled lamb physical properties.</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

Physical property characterization of chilled lamb by hyperspectral imaging and partial least squares regression

  • 期刊论文

  • 发表日期：

    2026年1月27日

查看全文
  • Xinxing Li,
  • Changhui Wei,
  • Buwen Liang,
  • Xuebing Bai

摘要

Traditional lamb quality testing relies on destructive physical and chemical analyses that are time-consuming and require strict experimental conditions. Hyperspectral imaging offers a non-destructive, rapid, and high-throughput alternative but faces issues such as data anomalies and weak predictions of nonlinear quality traits. Using chilled lamb, we established a spectral model for physical quality characterization based on hardness (g) and elasticity (g/sec). A second-iteration Monte Carlo sampling method was designed to improve data quality and reduce abnormal sample misclassification. We employed the statistical approach to filter irrelevant and redundant information from spectral data by using coupled t-detection and contribution measures. On this basis, a dimensionality reduction partial least squares regression (DR-PLSR) model was developed for quantitative prediction of lamb physical properties. The optimal model achieved an \(\:{R}_{P}^{2}\) of 0.986, an \(\:\text{R}\text{M}\text{S}\text{E}\text{P}\) of 23.986, and an RPD of 7.483 for hardness prediction, while the best elasticity model yielded an \(\:{R}_{P}^{2}\) of 0.812, an \(\:\text{R}\text{M}\text{S}\text{E}\text{P}\) of 0.016, and an RPD of 3.022. These results demonstrate the effectiveness of hyperspectral imaging combined with DR-PLSR for non-destructive assessment of chilled lamb physical properties.