Hyperspectral Imaging for Food Quality Analysis

Khan et al. (2020) reviewed Hyperspectral Imaging and its current status in the real-time quality control of food products.

What is hyperspectral imaging?

Hyperspectral imaging technology is a hybrid technology that combines imaging and spectroscopy to generate a spatial map of spectral variations. A three dimensional ‘hypercube’ of image data (x,y,λ) is generated using this technology by taking a series of two-dimensional spatial (x,y) images as a function of wavelength (λ) and super imposing them. Each image plane of the hypercube is composed of pixels and maps the light absorbance by the sample at a single wavelength,λi. Hyperspectral image data can be viewed in various forms: as a three-dimensional hypercube I (x,y,λ), as a two-dimensionalspatial image I (x,y) and as a collection of spectraI (λ) at pixel positions (x,y). Hyperspectral imaging data is of large volume depending on the instrument and camera pixels. The resultant data is big and needs good computing capabilities to process.

Hyperspectral Imaging in the Food System

Hyperspectral imaging has been used in a wide range of applications in the food industry for monitoring the quality of natural foods (e.g. fruits and vegetables, meat, cereals and nuts) and processed foods (e.g. milk powder, cheese, processed meat, coffee, food powders anddried fruits, etc.). Have a look at this YouTube video to see how it is used for quality monitoring of cheese.

Quality Monitoring of Natural Foods

Hyperspectral imaging (HIS) has been used for monitoring the quality of natural foods at the lab and industrial scales. This includes:

• Determining the origin of salmon fish
• Predicting meat quality in terms of various attributes (such as, tenderness, freshness, for detecting microbial attributes
• Determining various quality attributes of fruit and vegetables (e.g. apple surface defects, ripening of bananas and chemical components (moisture content, acidity, sugar content and soluble solid content))
• Evaluating the qualuty of cereals, e.g. rice quality
• Prediction of protein content in wheat
• Measuring internal chemistry (e.g. measuring glucose in potatoes, or protein and fat content in meats)

Quality Monitoring of Processed Foods

Hyperspectral imaging (HIS) has been used for monitoring the quality of processed foods. However, this has been mainly at lab scale. For processed foods, advanced machine learning and data mining algorithms are required for quality monitoring due to the complex relationship between processing and quality attributes

Hyperspectral imaging of processed foods in the lab have included:

• Determine the protein content of processed meat
• Determine the iron content in sausages
• Evaluating the quality of dried banana slices
• Discriminating milk powders of different quality and of different origin
• Detecting melamine in milk powder samples
• Determine starch content in various types of cheese
• Discriminate between different coffee brands

There are a number of knowledge gaps in relation to hyperspectral imaging for the real-time quality control of food that may be responsible for the limitied success of hyperspectral imaging in the food industry to date, especially for processed foods and in online applications. To date, hyperspectral imaging in the food industry is typically conducted as an off-line machine vision technique for quality testing to replace existing manual sensory tests. Overall, with more research into this still emerging field, real-time quality control in many industries is a likely future possibility.