Improving Flaw Detection with Machine Vision Systems

Introduction

Machine vision systems use cameras, computers and software algorithms to carry out inspection tasks that require precise and repetitive verification and testing in high speed, and one of the most prevalent machine vision applications has to be flaw detection.

Detecting flaws by employing machine vision systems is commonplace across a wide range of industries, including semiconductors, pharmaceuticals and automotive manufacturing, because machine vision systems not only lay bare all contamination, scratches, cracks, blemishes, discoloration, gaps and other imperfections undetectable by human vision, they also make available faster TTM and better resource optimization while at the same time lowering both the cost of ownership and the product fail rates.

Customer Requirements

A system integrator that specializes in smart, safe and sustainable machine vision technology learnt about Lanner while seeking hardware solutions for building an end user solution intended for ceramic tile inspection.

This flaw detection system would automate the most stringent and thorough check on all types of ceramic tiles for any kind of defect and for determining their patterns, colors, textures and eventually their values, grades, usability and sale prices. The top three most imperative requirements are first of all, a high-performance CPU; second of all, a rich I/O interface with support for at least 3840 x 2160 imaging resolution and third of all, a fast and generous GbE interface for speedy networking.

The CPU must be powerful enough for enabling tile inspection at a rate of up to 30,000 pieces of tile per hour while the I/O interface must be flexible enough for supporting the installation of multiple cameras and the connection to the factory network for remote management. Finally, the hardware and the resulting system must be reliable for 24/7 availability and should require the absolute minimal amount of manual maintenance and troubleshooting.

Lanner Solution

Lanner’s LEC-2580 is a compact, fanless box PC with integrated pluggable Intel Movidius accelerator engine. The LEC-2580‘s processing prowess comes from Intel®’s Core™ I series processors and its standout features include multiple serial ports, input/output and display interface that support ultra high resolution. The LEC-2580 and the LEC-2580P do also feature various Ethernet and PoE ports for greater networking capability.

Customer Benefits

The LEC-2580 is one of the solutions streamlined with Lanner’s configure-to-order manufacturing practice, meaning Lanner fully controls these solutions’ design and manufacturing process while Lanner’s OEM clients reap the benefits of having design flexibility and dynamics, resulting in significant reduction in development costs, support costs and time-to-market.

The LEC-2580 comes with Intel®’s 6th generation Intel® Core™ i7-6600U, i5-6300U, or i3-6100U CPU, up to 16GB of DDR3L memory and delivers outstanding image-processing performance for intensive graphic processing. Its four PoE connectors enable IP smart camera connections while the dual LAN ports can be utilized for serving data streaming, remote control and other networking purposes.

For connecting to factory peripherals like PLCs and robotic controllers, the LEC-2580 provides four RS-232/422/485 ports as protocols for peripheral coordination. Last but not least, the dual HDMI ports support up to 3840 x 2160 (at 24Hz) or 2560 x 1600 (at 60Hz) of resolution.

The LEC-2580 also leverages Intel’s industry leading embedded deep learning toolkit OpenVINO and Myriad X VPU’s for modular, low-power and cost-effective edge vision processing, not only delivering seamless hardware/software integration with pluggable Intel Movidius accelerator engine, but also making the LEC-2580 a developer-friendly and robust deep learning vision appliance for not only machine vision but also intelligent surveillance, traffic management, access control, retail and beyond.

Click here to learn more about using the LEC-2580 and Intel’s Open Visual Inference and Neural Network Optimization (OpenVINO™) toolkit for better integration and faster deployment when enabling AI capabilities at or near the edge.

Conclusion

After having successfully built and provided the flaw detection/inspection system using the LEC-2580, the said system integrator has raised the possibility of working with Lanner for building flaw detection/inspection systems using high resolution, high sensitivity image sensors/cameras, powerful frame grabber/vision processor hardware and proprietary vision software for flat panel glass inspection and rail inspection.

Improving Flaw Detection with Machine Vision Systems was last modified: September 2nd, 2021 by Jorge Peregrina