Quantification and visualization of meat quality traits in pork using hyperspectral imaging
1State Key Laboratory for Pig Genetic Improvement and Production Technology, Jiangxi Agricultural University, Nanchang 330045, China.
Meat Science
|December 1, 2022
Summary
Hyperspectral imaging accurately predicts meat quality traits and visualizes their spatial distribution. Adding texture data further enhances prediction accuracy for industrial applications in automated meat quality assessment.
Area of Science:
- Agricultural Science
- Food Science
- Spectroscopy
Background:
- Accurate meat quality assessment is vital for the food industry and pig breeding.
- Current methods often lack spatial resolution, limiting the evaluation of trait variations within meat samples.
Purpose of the Study:
- To evaluate the predictive potential of Visible/Near-Infrared (VIS/NIR) hyperspectral imaging for 14 meat quality traits.
- To assess the impact of incorporating texture information on prediction accuracy.
- To visualize the spatial distribution of key meat quality traits.
Main Methods:
- Collected large-scale VIS/NIR hyperspectral images using SpecimIQ.
- Developed predictive models for 14 meat quality traits using hyperspectral data.
- Integrated texture information to improve model performance.
- Utilized the best models to visualize spatial variations of Fat (%) and Moisture (%) content.
Main Results:
- Hyperspectral data alone achieved prediction accuracies (R²cv) of 0.60–0.70 for most meat qualities.
- Incorporating texture information improved prediction accuracy by 1.5%–16.4% across all traits.
- Successful visualization of spatial distribution for Fat and Moisture content was achieved.
Conclusions:
- VIS/NIR hyperspectral imaging is a promising technology for predicting multiple meat quality traits.
- Combining spectral and texture data enhances predictive capabilities.
- This technique offers potential for automated, spatially resolved meat quality assessment in industrial settings.
