Better control of food quality and composition with hyperspectral imaging

Food quality and safety are topics that raise a lot of attention these days. Consumers are highly aware of the effects of food quality and composition not only on person’s nutritional status but on the overall well-being. Ethical questions of food production become important, and on the other hand different type of allergies and intolerances are very common. It is not extraordinary that a regular consumer chooses a gluten free non-lactose organic meal as a regular working day lunch. In addition to increasing requirements from the consumers, food producers have a legal obligation to control the quality and safety of their products. Sophisticated analytical methods are required to support the development of new products that meet these consumer and statutory requirements, while also delivering an enjoyable sensory experience. In food manufacture, rapid methods are required to monitor and control production, to ensure consistent quality and to provide safety assurance. Thus the food production process control has to work with very short response times. Failure to respond quickly to Any faults in the production process may result in a waste product, lost sales, and could potentially endanger customer safety, all of which can escalate into big financial losses and health issues for consumers.

From lab to online

There are several technologies methods available to monitor the quality of food production processes. Unfortunately, many of them require laboratory analysis, which is labor intensive and may are very slow, the analysis might take hours or days or weeks from measuring the food sample to receiving the analysis result. Also, the sample is often destroyed during the inspection.

Benefits of hyperspectral imaging in the SWIR wavelength region for food applications

Shortwave infrared (SWIR) spectroscopy offers a fast and nondestructive method to inspect and monitor food quality and is well established in the food industry for laboratory and online use. This technique allows determination of the composition of samples (e.g. moisture content, protein content, and some ) as well as physical characteristics such as(e.g. particle size of ground particulate materials). Many relevant chemical bonds in food samples absorb light at particular wavelengths in the SWIR (900 – 2500 nm) range. These absorption features reveal the chemistry of the sample and can be detected by spectroscopy. The method is ideal for analysis of the average composition of homogeneous samples such as flour. For more complex products, The disadvantage of spectroscopy is that it only provides information on one single point in the sample at a time.

Hyperspectral imaging combines spectroscopy with imaging ability. This technique offers the possibility for simultaneous pixel by per pixel analysis of composite or heterogeneous foods. As an imaging technique, hyperspectral imaging doesn’t require contact with the food destroy or contaminate the sample, and with a good calibration model, the analysis result can be provided in real time.

Measurements of individual absorbance bands or calibrations for full spectra provide information on composition, which can be mapped within the image to measure the distribution of components such as moisture or fat. Hyperspectral imaging also offers a solution to complex, multicomponent product analysis which is difficult to do with non-imaging techniques.

Hyperspectral sensors provide information on hundreds of narrow wavelength bands. This provides data on multiple, overlapping absorbance bands for different chemical bonds, enabling calibrations to be developed for specific analytes of relevance for food production.While this is a major advantage in information amount compared to e.g. RGB imaging, it also sets a challenge. The user (or the creator of the application) has to know, which wavelength bands are of interest and are recorded for analysis, and which can be left out.

SPECIM, as a pioneer in hyperspectral imaging, has been providing hyperspectral sensors already for the last 21 years. During this time hyperspectral imaging has grown from highly scientific niche technology used mostly in airborne imaging and in the military to a highly interesting challenging technology for industrial use. Earlier hyperspectral detector technology was not mature enough to meet the requirements of online production as the recording speed has been way too slow. During the last few years the development has been tremendous, and today hyperspectral cameras can reach speeds of thousands of frames per second and illumination times less than milliseconds.

Campden BRI pioneers hyperspectral analysis services for food industry

Campden BRI is a UK based company whose history traces back to the year 1919 when it was opened as a Fruit and Vegetable Preserving Research Station. Today, Campden BRI is the world’s largest membership-based food and beverage research organization with 2,400-plus members in nearly 80 countries. Members include companies like Arla Foods Ltd., Kellogg’s, Coca-Cola, Heinz, and Nestlé.

Ten years ago Campden BRI decided to explore the opportunity that hyperspectral imaging provided to strengthen their food analysis methods and to expand their food imaging capabilities. “SWIR spectroscopy was already well established in the agri-food sector for the rapid analysis of foods and their ingredients. Hyperspectral imaging provides the opportunity to apply this approach to new applications and provides a unique way to measure, for example, the distribution of moisture and fat in complex food samples”, states Dr. Martin Whitworth, Principal Scientist at Campden BRI responsible for leading image analysis research.

For online production, short imaging time is one of the key requirements. Even for off-line analysis and product development, short imaging time is a critical requirement due to the strong external light source used to illuminate the sample: short exposure times are required to avoid damage such as drying or melting of samples.

Based on these criteria, Campden BRI decided to select pushbroom hyperspectral imaging due to the short imaging time it provides. Pushbroom hyperspectral imaging detects the full spectrum of a narrow image line at a time, and the full sample gets scanned with the production line movement. Since only a narrow line is imaged at once, imaging time can be reduced to millisecond level.

Campden BRI had a clear set of specifications in their mind when they started looking for a solution that met their criteria. They selected to go with a SPECIM SWIR camera. “It had to operate on 900 – 2500 nm, it had to be versatile for a wide range of product sizes, it had to do the measurement in a few seconds, and it had to be transportable for use in different production locations. We have been using this very same sensor now for nine years”, says Martin Whitworth.

Today, Campden BRI has been offering hyperspectral imaging analysis services for food industry already for a decade.

The system has been applied to a wide range of product types including bread, biscuits, grain, meat, fish, confectionery and fried products. A common application is to measure moisture distributions in products, either to study the effect of production conditions such as baking on the uniformity of final product moisture, or to measure changes in the moisture distribution over shelf life. This can be particularly important for products with multiple components of differing water activity, such as a low moisture product with a high moisture filling. Another application is to study the fat distribution in fried products.

Much of Campden BRI’s work is contract analysis for individual food manufacturers. In several cases, this involves the use of calibrations developed exclusively for those clients’ products. Many clients use the service to support product development work, to enable them to assess the quality of products from production trials. Samples are typically sent to Campden BRI’s laboratories, or the instrument can be transported to client sites for analysis of samples fresh from the production line.

In addition to providing information for product development, this also enables clients to evaluate the suitability of the method for planned online applications and to aid specification of the required instrumentation. The ability to apply the same approaches used in the laboratory to online applications was another key reason to select a pushbroom system.

Example applications for food analysis

Some applications of hyperspectral imaging in other markets use the spectral data to identify and classify features of different composition in an image. For many food applications, the full benefit of the method is a quantitative analytical instrument to measure the concentration of particular compounds. This is achieved by creating calibration models based on a comparison of hyperspectral images with reference measurements made by traditional methods for a series of calibration samples. There are plenty of different algorithms available (Partial Least Squares, Support Vector Machine, Neural Networks to name a few) to build these calibration models which map the desired parameters in the sample to hyperspectral data output.

Depending on the analyte, different absorbance bands will be of particular importance for the calibration. For example, crystalline sucrose has a characteristic absorbance peak at 1435 nm, lipid contains CH₂bonds with absorbance bands at 1724 and 1762 nm, and OH bonds in water molecules have several SWIR absorbances, including at 1925 nm. In some cases, qualitative assessments can be made using images at these specific wavelengths. However, best results are achieved using multivariate calibrations for a range of wavelengths, including for properties where the relevant choice of absorbance bands is unknown, or where there are multiple, overlapping bands.

After building the calibration model, it can be directly applied to the hyperspectral images of unknown test samples, or samples on the production line to rapidly map these parameters. Some examples are given below.

Moisture distribution in bread

Moisture distribution is an important attribute in many food products, which affects texture and conditions favorable for microbial activity. Moisture can change over shelf life and is therefore associated with product freshness. The distribution is not always uniform, but this is difficult to detect without an imaging technology.

The Figure below shows moisture distribution in a fresh slice of white bread, from a study by Campden BRI. First, a calibration was built, and then this model was applied to real samples. The purple and blue color in the image indicates low moisture content, while yellow and red color indicates high moisture content. It is easy to see the moisture level increases rapidly towards the center of the bread, while the outer crust has a very low moisture level.

Based on Campden BRI’s studies, hyperspectral imaging can be also applied to detect moisture migration in composite products, which are traditionally more challenging than one-component products.