Hyperspectral Imaging in food industry

A possible solution to the necessary revolution in the food industry

Dr Maria Ricci, riccimaria@gmail.com

START

Index

Context

Examples

Technique

Challenges

Information

SECTION 01

Context

Quality

Safety

Authenticity and Compliance

Minimal food waste

Quality

Safety

Authenticity and Compliance

Minimal food waste

• colour (actual colour and homogeneity),
• pH,
• carbohydrates content,
• fat and fatty acid content (and distribution),
• water content or water holding capabilities,
• protein content,
• texture (tenderness, juiciness, firmness) ,
• freshness (frozen, hours post mortem, maturity),
• morphological defects (bruise, insect damage, ...),
• smell,
• taste

• total viable count (yeast, moulds, bacteria, ..)
• specific parasitises or bacteria,
• microbial spoilage
• faecal contamination,
• unwanted chemicals

• classification of different varieties of same plant,
• identification of origin of product,
• identification of other organic contaminants (fragments of other plants or of different animals or fungi),
• identification of additives

• early identification of compromised food
• actual freshness of food (level of ripeness or post-mortem time)
• variation in time of total viable count

How do we normally inspect food?

Precise, rapid and objective inspection systems throughout the entire food process is important to ensure the customers' satisfaction. It is therefore necessary to find accurate, reliable, efficient and non-invasive alternatives to evaluate quality and quality-related attributes of food products.

Reference

• simple
• technically not challenging

• subjective
• time-consuming
• laborious
• tedious
• inconsistent

• well established
• precise
• chemically sensitive

• destructive
• incapable of handling a large number of samples
• consume a lot of time

• simple
• fast
• easily accessible
• cheap
• gives spacial information and distribution

• very limited chemical components information
• no detecting invisible defects (sub-surface)

• chemical information of the sample
• internal information of the sample (depending on how operated)

no information on spatial distributions, requires lot of sampling points or avaraging

https://www.sciencedirect.com/science/article/abs/pii/S1466856413000775?via%3Dihub

SECTION 02

Technique

Imaging

Spectroscopy

Spacial information is essential for not homogenous samples

physio-chemical information of the point of measurement

Concept of hyperspectral Imaging

Reference

Concept of hyperspectral Imaging

By integrating two classical optical sensing technologies of imaging and spectroscopy into one system, hyperspectral imaging can provide both spatial and spectral information, simultaneously. Therefore, hyperspectral imaging has the capability to rapidly and non-invasively monitor both physical and morphological characteristics and intrinsic chemical and molecular information of a sample.

https://www.sciencedirect.com/science/article/abs/pii/S1466856413000775?via%3Dihub

https://www.sciencedirect.com/science/article/abs/pii/S1466856413000775?via%3Dihub

Reference

Acquisition of hyperspectral images

https://www.ntno.org/v01p0369.htm

​

Plane scan or Wavelength scan

Line scan or pushbroom

Point scan or whiskbroom

Snapshot

interactance

reflectance

trasmittance

There are three different sensing mode, depending on the relative position of light source and detector:

Reference

External quality features are typically detected:

• size
• shape
• colour
• surface texture
• external defects

It is usually used to determine:

• internal component concentration
• detect internal defects

! ATTENTION: only relative transparent materials!

It is usually used to determine:

• concentration of components deeper in the material
• detection of defects that are deeper in the material (not directly at the surface)
• less surface effects than reflectance

https://www.sciencedirect.com/science/article/abs/pii/S1466856413000775?via%3Dihub

SECTION 03

Information

Reference

Bruised Tissue

Bruised Tissue

https://www.nature.com/articles/srep35679

Reference

https://www.nature.com/articles/srep35679

Reference

Hyperspectral image processing methods

Because the data volume of a hyperspectral image is usually very large and suffers from collinearity problems, chemometric algorithms are required for mining detailed important information. These are the typical steps of a full algorithm for analyzing hyperspectral image.

https://www.annualreviews.org/doi/10.1146/annurev-food-032818-121155

Reference

https://www.perception-park.com/detection-contaminants-rice

Reference

https://www.stemmer-imaging.com/en/products/series/perception-park-core-compile-studio/

SECTION 04

Examples

+ info

• discrimination of live rainbow trout that are on different diets
• differentiation of fish species and determination of fish freshness

• texture
• moisture
• pH
• frozen/not frozen
• type classification
• total viable count

• frozen/not frozen
• protein content
• water content and water retain
• fat content
• microbial spoilage
• total viable count
• identification of minced pork adulteration with minced pork jowl meat

• fat and fatty acids content
• total viable count
• water content
• tenderness
• protein content
• microbial spoilage
• colour and pH
• adulteration with other meat

• faecal contaminations
• microbial spoilage and specific pathogens
• water content

• measurement of quality parameters of olive oil (free acidity, peroxide index, moisture content)
• discrimination of olive fruits at different stages of maturation

• defects
• bruises
• contamination
• mealiness
• firmness

• identification of starch in fresh cheese
• classification of commercial Cheddar cheeses from different brands

• melamine in milk powder

• decay
• canker
• defects

• defects
• contamination

• Identification of different roasting levels in green tea
• Identification of species in herbal tea blends
• Identification of geographical origin of green tea

• classification of cocoa beans from different geographical origins

• discriminating the floral origin of honey

• maturity

• identification of coffee bean varieties
• classification of Arabica and Robusta

• identification of Millet and buckwheat flour in black pepper
• identification of Papaya seeds in powders and berries of black pepper

• identification of green pea adulteration in pistachio nut granules

https://www.annualreviews.org/doi/10.1146/annurev-food-032818-121155

https://www.mdpi.com/2673-7256/1/2/8

• defects

• bruises
• colour

• bruises

• bruises
• firmness

• moisture

SECTION 05

Challenges

Where does the devil hide?

• Redundancy of information renders the classification challenging to achieve: improvements on the classification algorithms in terms of accuracy and speed are required
  
• Quite dispersive: extremely high number of applications and possible parameters
  
• Difficult trade-off between efficiency and broadness of the applicability

• Possible lack of good reference data for classification purposes mainly for missing standards in sample preparation and data treatment
• There are little studies directly performed at the idustrial sites but rather in laboratories
  

Thanks for your attention!

Do you have any question?

https://de.linkedin.com/in/maria-ricci-94175152

riccimaria@gmail.com