The NIRWatchdog software platform combines several modules to provide the ability to identify the deviation of a material from the standard material in a given population. The system uses as an input reading from a Near-Infrared (NIR) Spectrometer, and through machine learning provides a prediction result determining if a product deviates from normal. For example, a common use case includes using NIRWatchdog on commercial packaging materials. If the quality of the material has degraded or is subject to fraud, the system would identify it as deviating from the standard range acceptable for its class, and flag it for further scrutiny.
The process begins with a scan using a Near-Infrared Spectrometer, a handheld tool that takes a rapid, highly informative, scan of any item. The technology is based on the Raman Effect using the near-infrared spectrum response. Our software platform (NIR Watchdog) gets an input reading from a Near-Infrared (NIR) Spectrometer and using the machine learning AI-algorithms it builds the molecular fingerprint of the scanned product.
We work in two stages - a training stage and an inspection stage. During the training stage, NIRWatchdog identifies the personalized NIR fingerprint of your product and trains the comparison algorithm. This model is used during the inspection stage to identify a deviation of the inspected material from the standard (which was trained) in a given sample population. The scan results are sent to a comparison algorithm, which determines with a certain probability whether the scanned product could belong to the trained samples’ family. Our system is frequently used to detect counterfeit or fake products and as a quality control measure both in-factory and after a product has entered the market.
We currently work with a major food producer with a dedicated focus on quality control. Our system is capable of rapidly identifying problems related to package quality and content, as well as adapting to existing quality control processes. The tamper-proof and trusted process coupled with our own proprietary machine learning methods reduces the detection time by multiple orders of magnitude.
Inspection data is stored in the encrypted database on a blockchain, containing scan data, digital signature, location along with the hash value of the data. All data is kept secure to ensure the integrity of the process.