Because quality plays a huge role in customer experience and brand loyalty, it has always been, and will always be, a critical KPI to monitor and manage.
While defect detection, early-stage warnings, no-fault forward and other such quality initiatives are already in the deployed and scaled stage, the adoption of new-age technology and advanced analytics has been very limited. Dependence on manual inspection and eyeballing persists to this day, resulting in enormous time lapses that are still not foolproof.
Computer vision technologies with basic pattern recognition have seen early adoption in fast-moving assembly lines. These AI-powered technologies detect defects like dings and dents, color patterns, gaps, and more, which may otherwise have been missed by the human eye.
Bristlecone armed a beverage company with a powerful label inspection solution, aided by computer vision and pattern recognition. It compares label position accuracy, orientation and quality (e.g., tears and scratches) against a standard set of labels provided by the customer. The solution recognizes bottles and brands, while aligning them with their respective labels.
Computer vision is also helping to ensure packaging quality for a motor oil company. We assisted with printing QR code labels and then aligning the QR code-based item counting and accuracy to the batch and box (multi-level QR code aggregation and predefined packaging). This enables product traceability and distribution visibility, while also attacking fakes and pilferage in the network.
There is still a set of quality parameters where computer vision alone may not be sufficient – for example, validating weld quality. Low-quality welding will cause metal defects over time, resulting in poor customer experience like leakages and structural damage. Looking at the root causes of the weld quality, it was determined that the welding tip had carbon deposition. The solution? An AI-powered computer vision solution that provides operators with timely alerts on welding tip cleanups, resulting in an improved and consistent weld quality.
So, what does it take to execute this type of solution successfully?
Data is key. You need to have data that is based on real-time availability, accuracy and reliability. For any advanced analytics, what’s needed is not just the real-time data, but also a substantial amount of historical data. In the absence of historical data, as is often the case, we capture the data and its behavior in real time for a period of 3-6 months, which provides the solution with a test bed. Armed with the data and continuous model training and testing, patterns begin to emerge. This is an investment, but one that has a higher return on maturity.
Solutions like these not only accelerate the quality control process, but also significantly improve accuracy and reliability by reducing human intervention.
What can AI-powered computer vision do for you? Contact us to schedule a meeting.