Hyperspectral Imaging in the Food Industry: A Review
Because of the inherent risks involved with human consumption, the food industry is highly regulated. The FDA and USDA have many regulatory requirements, some of which apply to all food production, and some which are commodity specific (e.g. seafood and canned goods). Some of the safety mandates include
· Good Manufacturing Practice (GMP) requirements
· Hazard Analysis and Critical Control Points (HACCP)
· Labeling requirements
· Banned substances
· Concentration limits
These hazards generally fall into three categories
1) Physical hazards (e.g. metal, glass, wood, stone)
2) Biological hazards (e.g. mycotoxins, botulinum toxin, salmonella, listeria)
3) Chemical hazards (e.g. pesticides, toxins, residual solvents)
Aside from safety issues, food producers have to deal with a number of quality issues to ensure customer satisfaction and maintain good brand image.
1) Sensory parameters (taste, smell, texture)
2) Routine parameters (moisture, fat, protein)
3) Complex parameters (ripeness, rancidity/spoilage, potency)
To be successful, food producers must monitor all of these safety and quality issues both quickly AND efficiently. As simple as that is to say, test methods that are both quick and efficient at identifying safety and quality issues have eluded food producers to this day. There is a constant struggle between the need to carefully monitor and the need to keep production running quickly and cost effectively. The two biggest contributors to this problem are
-Latency of monitoring chemical parameters
That is to say that the variation between and within food samples is considerable. Even the most well thought out sampling techniques struggle to account for this diversity in monitoring. Compounding this issue is the fact that testing methods for a majority of parameters are quite lengthy, to say nothing about their costliness. In the wake of these issues, food producers almost unanimously limit their quality control and safety monitoring to
a) Tests they’re legally obligated to conduct
b) Tests that have an appreciable return on investment
That isn’t exactly ideal. They wouldn’t limit their testing if they didn’t feel they had to. After all, these companies really do want to produce safe and quality products. But they’re limited by their resources, especially in this case, their testing methods, which include
1) Tests that are quick but have limited usefulness
2) Tests that are useful but not quick
3) Tests that are objective but don’t account for inhomogeneity
4) Tests that account for inhomogeneity but aren’t objective
To hash this out a little further – the types of tests that food producers can do quickly are fairly mundane. This would include things like metal detection and moisture detection. They’re important sure, but they give generic information that doesn’t address many potential concerns. Methods like chemical tests and chromatographic methods, in contrast, are exceedingly useful. They provide detailed information about parameters that are more indicative, and more specific to a given product’s quality. But with these methods the information comes at a price, namely speed. With respect to inhomogeneity, all the tests that provide objective analyses (e.g. chemical and chromatographic) are point-based. One sample will give you one result, a different sample, even from the same unit, will give you a different result. And the reason they differ is because the samples are inhomogeneous. Humans are quite adept at identifying and accounting for this type of variation within and between samples, but carry their own problems. Chief among them is the fact that they are by definition subjective. Translating human-based studies into action in a food production environment is therefore problematic.
What the food industry needs is a method that can at once provide accountability to inhomogeneity and also facilitate the addition of specific chemical parameters without contributing to latency in production.
Believe it or not that method does exist; in fact it has existed for more than a decade now. It’s called hyperspectral imaging. It’s an optical-based method that utilizes various sensors spanning the electromagnetic spectrum that can be tailored to analyze nearly any chemical parameter. With the advance of computing power and the incorporation of artificial intelligence, this method is quite rapid. That solves one of our problems. But hyperspectral imaging does one more, taking not just single-point measurements, but full data cubes that capture the spectral responses of each pixel across an image. That solves the other problem.
In summary, the food industry has two major problems with production quality control 1) inhomogeneity and 2) latency in testing chemical parameters. Hyperspectral imaging solves these problems and allows for a superior quality control regiment that will in turn produce better products and benefit those companies in the end.