Tips for choosing a 3D vision system

January 15, 2018 // By Nigel Smith
With four times as many as colour receptors as humans, the Mantis shrimp has the most impressive eyes in nature. Manufacturers have long relied on human vision for complex picking and assembly processes, but 3D vision systems are beginning to replicate the capability of human vision in robotics. Here are the three rules to live by when choosing a 3D vision system for manufacturing.

Rule one: Abandon CAD

Advanced 3D vision systems are a stark contrast to the vision software of manufacturing’s past. Many existing systems still require professional CAD programming to ensure the robot can recognise shapes. However, even after programming, this software can have difficulties recognising multiple items at once.

A common application for vision systems is removing and sorting items from a bin. While CAD-based systems can identify items in a bin, the challenge is recognising the position of each item when presented in a random order — let alone determining the best method for the robot pick them in. 

Advanced vision systems eliminate this problem by using passive imaging to enable the robot to automatically identify items, regardless of their shape or order.

Toshiba Machine’s vision system, TSVision3D, for example, uses two high-speed cameras to continuously capture 3D images. Using intelligent software, the system can process these images and identify the exact position of an item. This determines the most logical order to pick them up and does so with sub millimetre accuracy, with the same ease as a human worker.


Rule two: Mimic human perception

Deploying a robot for bin-picking isn’t advantageous if the robot cannot identify the edges of the bin. Considering the speed and strength of most 6-axis robots, hitting the box sides could easily halt production or damage the product. 

Some manufacturers believe that motion stereo systems can effectively imitate a human’s perception of an item. Motion stereo systems use one camera, usually mounted on a robotic arm, to enable the system to move and take two or more photographs of an object. However, these systems require absolute precision as even the slightest movement can cause disparities in data and skew the measurement.