The inside story

Citation

Loughlin, C. (2000), "The inside story", Sensor Review, Vol. 20 No. 3. https://doi.org/10.1108/sr.2000.08720caa.001

The inside story

The inside story

Our theme for this issue is "Tomography and 3D imaging" and although most dictionary definitions state that tomography is the x-ray imaging of a slice through an object, the field of tomography now encompasses various electrical imaging techniques based on resistance and capacitance.

How do you arrange six matches, without overlapping, to form four equilateral triangles?

2D machine vision systems are now widely accepted in industry and business is flourishing. Less common are systems that provide 3D imaging of parts or scenes, although in many cases the use of 3D would actually make tasks easier rather than more complicated as might be anticipated. The third dimension is all too easily overlooked.

Those struggling with the above riddle only need to think 3D and consider a three sided pyramid to see what I mean.

To provide a more industrial example, consider a part emerging from some parts feeder. It could be a bowl or linear feeder or simply a series of conveyor belts. If the part concerned can only land in a limited number of stable states then it is quite likely that a measurement of the height of the highest point on an object would reveal which one of the stable states it was in. This alone may be sufficient to allow sorting or grasping to occur.

The recognition of an object and the determination of its orientation is frequently determined by the location of the objects extremities and/or internal structures, and this is where 2D can let us down. For example there are numerous algorithms for the location of edges within 2D grey scale images. The Sobel operator is one of the oldest and simplest but even the very latest, most sophisticated procedures are never 100 per cent reliable. This is because they all rely on changes in grey scale to locate the edge. While it is often the case that this will locate the edge, the change in grey scale is very much a secondary consequence of the edge being there, and has nothing to do with the physical shape that fundamentally defines the edge's existence.

Edges are the intersection of planes and surfaces, and only a 3D map of an object can truly define the location of these same edges. The resolution required of the third dimension may not be too technically demanding and in many cases it could be argued that you are better off with medium quality 3D rather than high resolution 2D images.

In the early days of machine vision technology advanced from low to high resolution and from binary to grey scale and even colour images. Backlit tables and pristine white conveyor belts were the norm - no wonder that industrial imaging has taken so long to really get moving.

I shudder to think that a large part of my PhD thesis was concerned with the application of 32 x 32 binary images. And consider that it will not be too long before we all view 2D grey scale vision systems with the same patronising sympathy that doubtless any current research student would bestow on my own past efforts.

Clive Loughlin