Image credit: DigiTimes
Ensuring quality control is critical for all companies, but for contact lens manufacturers it can be particularly problematic. Where most companies can inspect random samples for quality assurance, contact lenses are highly customized and the material is transparent, making it nearly impossible to inspect randomly or by machine. As a result, the lenses are typically inspected by humans, which is a slow, error-prone process that creates a bottleneck in the production process.
When humans manually inspect all lenses, at a rate of about 4,000 lenses per day, it can result in a high rate of missed or false flaw detection. Adding to the challenge is the transparency of contact lens materials, most of which are created from a silicone hydrogel. Typically, machine vision automated optical inspection (AOI) relies on fixed geometric algorithms to identify defects, but capturing quality images of highly transparent items is difficult and results in poor defect detection rates.
With consumer demand for contact lenses growing at an annual rate of 5 percent, manufacturers need a more efficient quality control process. LEDA Technology, a software and services company that serves the optical industry, has tackled this problem. Using the Intel® Distribution of OpenVINO™ Toolkit and ADLINK’s artificial intelligence (AI) machine vision technology, LEDA created a highly efficient and more effective automated inspection solution for contact lenses.
A Smarter Inspection Process
LEDA’s AI contact lens inspection solution is built on ADLINK’s AI machine vision technology. The smart inspection solution more accurately detects lens flaws, including burrs, scratches, tears, foreign particles, and bubbles. Based on the ADLINK EOS-i6000-M Series GigE vision system for the edge, it supports deep learning and improves AI by collecting data from inspection devices. Because the AI gets smarter as it collects more data, the inspection data continues to improve with use.
ADLINK’s EOS vision system is equipped with an industrial GigE Frame Grabber technology camera interface, which provides improved image and video quality from industrial GbE cameras. The image capture app also can connect to other equipment and the cloud, enabling software partners to deploy new solutions more quickly.
LEDA’s system uses the 9th Gen Intel® Core™ i7-9700E processor and four add-in accelerators based on Intel® Movidius™ Myriad™ X VPUs, optimized by the Intel® Distribution of OpenVINO™ toolkit. The inference engine enables it to balance workloads and improve efficiency, and heterogeneity support allows execution across the cloud, local servers, and edge AI vision devices.
Focus on Quality Control
LEDA’s automated machine vision inspection system has drastically improved quality control. Instead of inspecting 4,000 lenses per day, the machine can scrutinize 50 times that amount. Defect detection has also jumped from a dismal 30 percent accuracy rate to 95 percent, according to LEDA. The AI-enabled solution also maintains inspection logs, which provides customers with an audit trail and additional visibility into the manufacturing process.
In one scenario conducted by LEDA and ADLINK, the deep learning AI solution was put in place alongside human inspectors, and data was collected concurrently. According to internal data, within three weeks they were able to convert the previously used machine vision system and parallel computing platform with the Intel® Distribution of OpenVINO toolkit.
The smart AI machine inspection solution will allow contact lens manufacturers to increase productivity, improve lens quality, and reallocate resources to more profitable tasks. The potential for increased profits looks good to any company.
- For more information about the solution created by ADLINK and LEDA Technology read this Intel Solution Brief.
- Learn more about ADLINK’s EOS-i6000-M Series of AI Machine Vision devices here and its lineup in the Intel® IoT Solutions Marketplace.
- Find out more about LEDA Technology here.