January 28, 2013      

Typical robotic camera recognition is a time consuming process that can be improved by using a more human-like process , according to Joe Cyrek, vice president of Recognition Robotics.

Machine vision will look at an object as it comes into its field of vision, calibrate the pixels and the distance, and then will use that information to recognize the next object, Cyrek explained. That?s a time consuming process that depends on the object to be recognized coming through the same way every time.

Calibration is a time-consuming process that depends on the object to be recognized coming through the same way every time.

However, if the object, whether a candy bar or a large part for an automobile, comes through at a slight angle, the typical 2D robotic camera won?t recognize it.

As a result, much of the money companies invest in robotic cameras is to ensure that objects come into the field of vision at the same angle every time.

?That precision is expensive,? Cyrek said, explaining that much of the expense of an automotive assembly line is the guidance system on the conveyor belt.

Recognition Robotics?s CortexRecognition visual recognition and visual guidance software, by contrast, uses a single two-dimensional camera image and proprietary recognition-based algorithms to record information much like the human eye and brain.

Once the user records and stores the image through point-and-click teaching, the software can recognize the scanned object in space from the x, y, and z positions as well as rx, ry and rz (roll, pitch and yaw) without the need for a calibration system.

Works with any industrial robot and can learn a large number of objects

CortexRecognition can work with any industrial robot. It can learn a large number of objects then recognize and locate any of the learned objects regardless to its presentation in the visual field of the camera.

Robeye is Recognition Robotics’ application of the CortexRecognition software into a fully functional 3D robotic package.

visual cortex

Robeye works with any robot to simplify robotic guidance. The application can eliminate the need not only for the guidance system on a conveyor belt, but also for the conveyor belt entirely, Cyrek said. For example, Robeye has been used to recognize automotive structural parts directly from a rack without the need to put them on a conveyor belt first.

Users install Robeye by mounting a single camera with a gigabit Ethernet cable connected to a vision controller.

According to Recognition Robotics, it takes about five minutes to teach a new model the vision system.

Robeye can also save on personnel costs, because it eliminates the need for operator intervention to re-position items, while also eliminating the need for a robotic vision expert to calibrate the system, according to Cyrek. ?It?s an affordable eyeball.?

Widely used by car makers; now targeting bin picking and packaging industry

The product is already in use with auto manufacturers, including Ford, Volkswagen, Chrysler, Mercedes, Maserati and Fiat; Cummins Engines, as well as at Yaskawa Motoman and Comau, Inc.

Comau, a Southfield, Mich.-based provider of integrated robotics and other automation technology, had an exclusive license with Recognition Robotics for Robeye from 2008 and 2011, and is still one of the company?s prime customers, according to Cyrek.

Current Robeye guidance applications include pack loading, picking from a conveyor, rack loading, hemming, assembly, tire inflation and deflation and piston insertion.

The company also sees Robeye as ideal for bin picking because it can quickly identify the jumbled container items without first laying them flat.

Robeye is working on a simplified version of the device packaging applications which would work on four degrees of freedom (x, y, z and rz) rather than six.

This would still meet the needs of most in the packaging industry while also costing significantly less than the Robeye version used today in automotive manufacturing, Cyrek said.