Machine Vision with a CMOS-Based Hyperspectral Imaging Sensor Enables Sensing Meat Freshness.

Area of Science:

• Computational sensing
• Machine vision
• Spectroscopy

Background:

• Quantifying sensory properties of materials for consumer electronics is challenging.
• Hyperspectral imaging (HIS) captures both spatial and spectral data simultaneously for non-destructive remote sensing.
• Machine learning (ML) techniques can efficiently process complex hyperspectral data.

Purpose of the Study:

• To develop a measurable physical quantity, the freshness index (FI), for assessing meat freshness.
• To integrate HIS and ML for advanced food inspection applications.
• To demonstrate the potential of this system in consumer electronics like refrigerators and smartphones.

Main Methods:

• Utilized a hyperspectral imaging system (HIS) for data acquisition.
• Employed machine learning (ML) techniques, including linear discriminant and quadratic component analyses.
• Defined the freshness index (FI) based on meat fluorescence, correlating it with bacterial density.

Main Results:

• Successfully converted meat freshness into a quantifiable freshness index (FI).
• Demonstrated efficient processing of hyperspectral data using ML.
• Showed that the FI can be estimated from hyperspectral data for unknown freshness states.

Conclusions:

• The HIS integrated with ML acts as an 'artificial eye and brain' for advanced machine vision.
• This computational sensing system offers advanced sensing versatility for hyper-personalization and customization.
• The developed method provides a non-destructive, efficient approach to assessing meat freshness.