3D machine vision system targets highly reflective objects
Helmee Imaging, a Finnish start-up, has unveiled a 3D machine vision system geared to speed up the quality control of glossy and highly reflective objects. Based on a combination of structured lighting and stereo imaging, the system measures how the surface distorts predetermined illumination patterns, in place of traditional surface imaging, noted the company.
While camera systems with projected patterns and 3D image reconstruction are often used in the manufacturing industry for the inspection of mat objects, so far highly reflective objects could only be inspected manually, and sometimes quite approximately. This is an issue for many metal or ceramic products with coated or polished surfaces, such as tableware, faucets, artificial joints, optical components or bearings to name a few.
The system is capable of simultaneous measurement of the object's surface quality and 3D shape, with a typical spatial resolution of 0.1mm (in X-Y) and down to 1um in the Z-direction according to Helmee Imaging's CEO, Matti Saarinen who shared more technical specifications with EE Times Europe. This gives plenty of resolution to identify even the smallest bumps or cracks. The announced scan rate is around one second per part.
The patent-pending system is based on structural lighting and stereo calculation, with RGB lasers driven through micro-projectors to generate the illumination patterns and two to three cameras that record how these patterns behave on the object's surface.
Because the objects being inspected are highly reflective, the patterns are not directly projected onto the objects surface, but instead onto a semi-transparent optical dome covering the object. The dome bends the light rays and also diffuses them a bit.
Left, a prototype showing the three illumination sources, the translucent dome with projected patterns, and one camera mounted on top of it. On the right, reconstructing the object's surface through the pattern's light rays reflected by the object.
"Basically we create 80 million triangles on the object's surface. Two points of a triangle are known, the camera and the projector-dome point. Using these points we are able to calculate the surface point of the object and integrate the whole surface and 3D-shape of the object," explained Saarinen.
"The algorithm that calculates the object's properties from tens of images is compute-intensive. But then we can obtain reference images that can be used for actual inspection algorithms (contours, blobs, thresholds, texture analysis, etc.)," he added.
So when asked if there was any reflection artefacts to be removed digitally through computation, Saarinen answered logically: "In fact, we don't try to remove any artefacts because the system works best on mirror-like surfaces, and the more the object reflects light, the better."
There are two major parameters that need to be controlled with this optical inspection system, the dome's transparency and transmittance which we can vary from 30-50 percent depending on the dome used (material and size) and the pattern's resolution. These are parameters that would be set with a reference sample to optimize the inspection in the production flow.
Helmee Imaging is working with a German tableware manufacturer whose products come with a highly reflective glazing, but in principle, the system could find many other industrial applications.
- Julien Happich
EE Times Europe
|Related Articles||Editor's Choice|