An overview on 3D Vision by Machinery Automation and Robotics’ own expert Roland Painter. This episode provides an overview of what exactly 3D vision is, why and when it is used, various acquisition options, profiling methods and much more. Roland’s clear experience on the vision subject matter combined with his lighthearted and approachable manner make this an easy to watch overview for the technically minded and those with a general interest in the field of vision, automation and robotics.
What is 3D Vision
Loosely defined 3D vision comprises of any vision system where in addition to standard two dimensional images the distance from the camera is also determined across the field of view.
Why use 3D vision
For some applications the third dimension is required to provide desired information. This could include the segmentation of objects from a similar background extracting the shape detail of dark objects, and identifying information surrounding the location or orientation of an object.
The sensor utilised within this MAR Ask the Expert demo is a commercial structured light sensor. There are however various options available to accommodate 3D Vision, a few of these include:
- Laser profiling: Utilising the triangulation of a laser sheet of light to capture, display and record desired data.
- Stereo vision: Two cameras recording same scene from a separated distance, with images overlayed to find disparity and calculate distance.
- Structured light: Projecting a known pattern to allow for the calculation of surface information based on the patterns deformation.
- Time of flight: Measures the time taken for a light pulse to be reflected from the scene.
When to use 3D Vision
The addition of a third dimension is best utilised for any application that cannot be simply solved with greyscale or colour vision alone. The addition of a third dimension is also capable of simplifying sensing and identification process for tasks that may traditionally use more complex analysis of two dimensional information.
Some of the options available in processing the image from a 3D camera include:
- 2.5D image processing: Makes the image pixels either in coloured or greyscale according to the distance allowing for the application of thresholding on distance or 2D vision functions.
- 3D point cloud: Opens up a few more opportunities with the ability to reorient point cloud, apply transformation functions, match 3D patterns, segmentation and filtering in 3D, and if needed render the image to 2D
Calibration & Correction
Often a 3D vision system will need to be calibrated to real world coordinates. This involves the use of a calibration target and significant 3D maths – vectors, matrices and quaternions.
There are a vast and growing number of applications for 3D Vision, including but not limited to:
- Robot guidance – From tasks such as picking / packing through to mobile robot guidance and mapping.
- Quality control – Identifying defects in products or shapes, or even conducting dimension checks.
- Measurement – Able to measure dimensions, including volume, allowing for things like constant volume slicing control
- Object identification / classification – Capable of allowing the identification and classification of a presented object.
For more information on 3D Vision or other robotic and automation enquiries, ask the experts, ask Machinery Automation and Robotics.
Have a question? Email: firstname.lastname@example.org