Image credit: Mariner
Integrating artificial Intelligence on the manufacturing floor improves product quality and productivity. AI can be used to automate processes, enable robotic assistance, and monitor equipment and products. Quality control, specifically defect detection, particularly benefits from AI-based machine vision systems. These solutions use digital images to train AI models to find defects faster than human employees can.
In some cases, however, machine vision systems aren’t living up to their potential. They might be able to identify a discrepancy, but they can’t classify the specific problem. Is there a true defect? Is there debris on the product, or perhaps the lens? A manufacturer won’t risk its business by ignoring flagged flaws, so companies overcompensate by tagging all potential defects then having human inspectors follow up on each one to verify actual defects.
That is a slow, expensive process, and now, it’s an unnecessary one.
Next-gen Machine Vision
Mariner has created the Spyglass Visual Inspection solution, a new machine vision system for defect detection. The SVI solution applies deep learning artificial intelligence, Internet of Things, and cloud computing technologies to identify defects during manufacturing. It lowers false positives, improving quality and reducing waste on the production line.
“Not only are we very accurate at identifying and classifying defects, we’re also helping companies to eliminate defects in the first place,” says Phil Morris, co-founder and CEO of Mariner. The Spyglass system, he adds, “will alert personnel when defects exceed a designated threshold. It can tell when the production line is running in a suboptimal manner.”
SVI is camera-agnostic, so it can be deployed on top of a manufacturer’s existing machine vision solution. Using the images captured from those cameras, Spyglass will take over the decision-making processes from the original solution. The images are sent to an edge server, which forwards the telemetry data to the cloud for analytics and retraining of the AI model.
Train and Retrain
Many AI-based defect detection solutions use pre-built libraries to train the model, but they tend to fall short, says David Dewhirst, vice president of marketing for Mariner. “We have data scientists on staff to work with the customer to get it right,” he says.
Human inspectors also can tag images to improve the AI training, just as they would when training an employee. The frequency of the retraining depends on the customer, product, and number and type of flaws detected. It can even be impacted by environmental factors, such as changes in lighting. When certain thresholds are met, SVI automatically triggers the retraining, and the new AI models are sent to the edge devices.
In addition, SVI can perform root cause analysis. It communicates with other systems on the line to identify the source of the defect. It can isolate an errant process or machine and send a notification, so an employee can view and remedy the problem before additional defective items are manufactured.
Augmenting an existing machine learning system is the easiest, fastest method of deployment, but Mariner has channel partners to fulfill other deployment scenarios. In new deployments, Mariner works with Ohio-based system integrator North Coast Technical Sales to provide and install cameras and lighting that enable Spyglass to operate optimally. Arrow provides the Dell edge servers, powered by Intel® Xeon® processors, and has expanded Mariner’s global reach.
Mariner is an Intel Gold Partner and Spyglass Visual Inspection is available as an Intel® Market Ready Solution.
- Learn more about Mariner.
- Find out more about Arrow Electronics.
- Learn more about North Coast Technical Sales.
- Discover more about the Intel Xeon processors.
- See the Mariner Spyglass in the Intel® IoT Solutions Marketplace.
- Listen to the podcast discussion on defect detection technology led by Erin Barrett, Chief Revenue Officer at Eigen Innovations.
Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries.