Acrovision's Deep Learning AI Vision System at Toyota

At Toyota quality is key. So, when improvements were identified in their existing manual inspection method, Acrovision helped Toyota develop a new system using the latest Artificial Intelligence vision technology. Together, a concept was formed that would reduce costs and improve efficiencies.
It was clear that traditional “Rule-Based” vision technology could not replace this human decision-making process, hence AI or Deep Learning vision became the technology of choice.
The Smart Auto Check System (or SMACS) consists of LED lighting, high resolution HD cameras feeding images to a PC via Network Switches and GigE Capture Cards. The heartbeat of the system is controlled using Cognex’s Deep Learning software driven by a high-end PC with Nvidia graphics card. Our custom C sharp front-end interfaces with the Shop Floor Applications.
Introduction of the system starts with the training phase, sample images are used to teach the software - to locate, analyse, classify and read anomalies and defects… even with variations. The solution doesn’t use a rigid set of rules, but is example-based, modelled on human learning where a variety of good and bad examples are ‘shown’ to the system, and the classification is validated. This cycle is repeated and refined, until the AI decisions are consistent, reliable, and proven against manual inspections.
Once deployed, if SMACS detects any anomalies, the vehicle is taken off the line for repair. Data is fed into the quality KPI reporting systems, allowing production quality to be monitored along with the system’s performance.
Acrovision’s first AI system integrated into Toyota’s production includes 17 cameras confirming the specification for each vehicle is correct, by identifying components and checking for defects. Since its introduction, there has been a significant reduction in the inspection time.
But we haven’t stopped there. The second SMACS is now in the training phase, performing 60 inspections around the each car body, using an array of 12 cameras.
By harnessing the power of deep learning that tolerates naturally occurring variations, SMACS detects defects with unprecedented speed and accuracy.
From initial consultation, through to development, programming, installation and commissioning, Acrovision are proud to have provided Toyota with such a successful, robust and innovative solution.