Ease of use and processing power have also dramatically improved. Previously, manufacturers did not want to roll out their own vision system. "Now the smart camera basically is the vision system," Singleton says. "It's easier to implement, especially for application-specific solutions." With these ready-to-go systems, the use of vision is more widespread. In many vision applications, "You don't have to have an automation engineer, vision guru or systems integrator," says Singleton, who maintains that there will always be a need for integrators in complicated implementations. Now systems are usually set up so the typical manufacturing person can implement vision-they know the process, what needs to be done and what the vision system has to interface to.
Now companies are introducing neural network technology-self-learning systems that are specifically tailored to target particular tasks. "It's been hyped for years, but they didn't really work," Singleton says. "In the past, general purpose systems didn't really know what needed to be done." Companies would pay a lot for the system and wanted it to handle many different tasks. For example, they would try to bundle assembly verification and color inspection together, which didn't work. Now companies focus on specific applications.