Toyota is expanding its use of AI vision through Deep Learning which automates vehicle checks

Acrovision’s Deep Learning inspection system has improved the overall performance of checks and also delivers new inspections when required for Toyota Motor Manufacturing UK.

For this project, Acrovision was asked by Toyota to provide an automated inspection system to replace, automate and enhance the current manual inspection process to improve quality reliability and efficiency.

After an initial first stage of what Toyota call SMACS (Smart Auto Check System), the benefits were so successful, including significant reductions in inspection times, that this second phase has now been implemented.

The initial requirements were as follows:

• An automated inspection system to replace current manual inspection
• For the data to communicate directly with Toyota’s tracking system
• The system to hold up to 50 vehicle body numbers in advance of inspections
• That the use of AI/Deep Learning technology would benefit these inspections
• The ability to install the system during limited line-availability

The new solution consists of 12 x GigE PC cameras with external lights carrying out over 60 inspections on each vehicle. As the car moves along the track, the cameras are triggered multiple times to acquire images of the various inspection areas. These images are communicated back to the central PC for processing and decision making.

All relevant data is fed back to Toyota’s tracking system for logging of results/defects. This triggers any faulty vehicles to be removed at the next checkpoint for corrections to be made.

Acrovision recognised that replacing a human decision-making process could not be achieved using traditional ‘rule-based’ camera technology. The development of AI or Deep Learning vision has enabled vision inspection solutions to work more like a human brain by using neural network training of good/bad examples.

Acrovision therefore designed a solution using a combination of Cognex’s VisionPro PC Vision platform and ViDi Deep Learning tools, to create a bespoke offering with a customised front-end. This communicates directly with Toyota’s host tracking system via broadcast messaging providing the required individual vehicle information, which in turn determines what inspections are required for that particular vehicle.

As a result of this second phase, Toyota now has a fully working solution which automates its vehicle inspection checks.

The solution was also designed to enable Toyota to add new or edit existing inspections and, says Paul Cunningham, managing director of Acrovision, in-depth training was provided to facilitate this.

Due to the success of this new installation, Toyota has further improved the overall performance of its inspection processes and subsequently has used AI in other areas of the business, including the use of Deep Learning Smart camera-based solutions.

Tel: 0845 337 0250
Website: www.acrovision.co.uk