AI-Assisted Hyperspectral Imaging Over NIR Spectroscopy in the Analysis of Food Safety and Quality
The Advantages of AI-Assisted Hyperspectral Imaging Over NIR Spectroscopy in the Analysis of Food Safety and Quality
Picture Credit: Science Photo Library
Near-Infrared Spectroscopy, aka NIR, is an optical-based test method that uses a spectrophotometer to analyze the light spectrum reflected off a test sample when it’s exposed to infrared light. The resulting reflectance is characteristic of the chemicals within a sample, giving information about its chemical composition.
NIR Spectroscopy is routinely used in the flow of food steps in a company’s Hazard Analysis Critical Control Point (HACCP) Plan for quality control and to stay within FDA and USDA mandated standards that specify if a chemical constituent is allowed to be present, and if so, at what concentration. For these purposes NIR is used:
Commonly to test broad routine parameters such as moisture, fat, and protein
Can also test specific chemical components like caffeine, lactose, vomitoxin, etc.
NIR is thought to have benefits over chemical tests that lend themselves to being used for quality control and hazard analysis in a process environment
Little to no sample prep
High speed - when compared to chemical tests
Low cost - because it doesn’t use consumables
But NIR as a method for quality control also has a number of disadvantages
Point-based - forcing sampling techniques to make up for a small amount of samples representing the whole
Fixed wavelength range - limiting parameters and making certain chemical analyses impossible
Low Accuracy compared to chemical tests
AI-Hyperspectral Imaging (AI-HSI), like NIR spectroscopy, is an imaging-based technology that gives information about the chemical composition by measuring its interaction with light.
However, AI-HSI has a number of advantages over NIR spectroscopy in quality control and the analysis of food safety issues
Larger sample sizes - giving higher throughput of material and better representation
2D spatial distribution of parameters - telling you not just what, but where
Expanded list of parameters - both because the wavelength range can be adjusted and because it’s able to make determinations on datasets NIR can’t
Better accuracy - produces better results from the same data due to special algorithms
Built-in feedback loop - Learns from samples to improve future analyses
Less data points - Requires less data to make the same statistical conclusions
Because of these advantages, AI-HSI is rapidly replacing NIR as a superior method of analysis and is gaining a strong foothold in the food industry.