|
|
A research project has been established within the Faculty of Engineering UWE, to develop innovative techniques for the visual inspection of complex surfaces.
In many cases industrial surface inspection is still undertaken manually
by trained staff involved in costly, tedious and time consuming operations.
This absence of automation may be attributed to existing methodologies, which
are often unable to cope with a wide variety of products and defects, yet
a continued reduction in computing costs would suggest that automated industrial
inspection has potential as a cost effective alternative. The wider application
of automated surface inspection would seem to offer several advantages, including
reduced labour costs, the elimination of subjective judgment, and the creation
of timely statistical product data.
The research effort has been aimed at generating novel techniques for the quality inspection of manufactured components, with difficult or complex surfaces, for which conventional machine vision techniques may prove inadequate. This includes the detection of surface defects such as scratches, erroneous indentations or protrusions, present upon smooth and textured surfaces, for both nominally flat and more complex three dimensional forms. The difficult problem of the inspection of surfaces which include a coincident chromatic pattern, or motif, tending to conceal underlying surface topography, has in particular been addressed. Surface chromatic patterns may be either regular, pseudo-random or stochastic. Typical applications include the inspection of decorative ceramic wall tiles and china products, polished natural stone surfaces, used as decorative cladding, and a range of painted or transfer printed manufactured components. The subsequent automated classification and quantification of defects is also seen as an important consideration in the realisation of automated manufacturing closed loop process control.
A strong emphasis has been placed upon identifying techniques with potential
to facilitate the wider industrial application of machine vision inspection,
by simplifying practical implementation, through for example, utilising a
fixed generic lighting configuration, and incorporating a tolerance to variation
in object position and orientation. These attributes are seen as a desirable
prerequisite to the greater implementation of automated surface inspection,
by eliminating existing restrictive environmental structuring, often typical
of vision based inspection systems. It is also desired to avoid the
need to undertake an initial training stage, in practice an expensive and
time consuming operation, and often necessary in order to isolate topographic
defects while in the presence of complex and irregular surface colouring.
An innovative technique has been developed for the acquisition and separation of surface topographic detail, from coincident surface coloured patterns. The selection of sample images shown below serve to demonstrate the process, for which conventional vision inspection techniques might fail to detect the hidden surface flaws. The images show how a concealed underlying surface flaw can easily be revealed by automatically removing any coincident surface chromatic pattern.
Transfer Printed Bathroom Tile
Camera Acquired Image Revealed Surface Defects
Polished Granite
Camera Acquired Image Revealed Surface Defects
Defects in Particle Board
Camera Acquired Image Revealed Surface Defects
A domain mapping technique is next performed, in order to classify and quantify the isolated three dimensional surface aberration. The recovered surface may also be displayed in relief, as in the sample images shown below.
Recovered Defects in Surface Relief
Recovered Complex Relief
The new approach also has application for the inspection of complex topological features which may be obscured by a coincident coloured pattern. The example below represents a particularly difficult surface inspection problem, and shows a non-planar surface possessing a pseudo-random chromatic pattern. The surface incorporates a regular three dimensional topographic surface relief pattern, concomitant with a pseudo-random albedo pattern. During inspection it is necessary to ensure the integrity of the concealed topographic form. From the acquired image it can be seen how the chromatic pattern tends to obscure the topographic pattern, making inspection using conventional image analysis highly problematic. The figures show how the new technique is readily able to separate the surface colouring from the obscured surface topography.
Acquired
Image (concealed surface topography)
Revealed
Surface Topography
It is important to appreciate that in each of the above examples, no initial training was required, the same lighting and camera configuration was used, and the object pose was not constrained.
We are hoping to attract government funding to continue this exciting work, and are keen to develop interested industrial and academic contacts. We believe the technique to have particular application for the detection of surface defects, upon a wide range of manufactured products.
M. L. Smith, G. Smith, T. Hill, F. Meyer, An environment for off-line configuration and programming of a vision based inspection system using CAD data, The Caledonian International Engineering Journal, Vol. 1, pp.31-39, 1996.
M. L. Smith, T. Hill, G. Smith, Surface texture analysis based upon the visually acquired perturbation of surface normals, Image and Vision Computing Journal, Vol. 15, No. 12, pp.949-955, 1997.
M. L. Smith, G. Smith, T. Hill, Gradient space analysis of surface defects using a photometric stereo derived bump map, Image and Vision Computing Journal, Vol. 17, No. 3-4, pp.321-332, 1999.
M. L. Smith, The analysis of surface texture using photometric stereo acquisition and gradient space domain mapping, Accepted for publication in: Image and Vision Computing Journal, 9 pages, January 1999.
M. L. Smith, R. J. Stamp, The automatic visual inspection of textured ceramic tiles, Submitted for publication in the journal: Computers in Industry, January 1999.
For further information, contact
Dr. Melvyn Smith
Faculty of Engineering
University of the West of England
Frenchay Campus
Coldharbour Lane
Bristol BS16 1QY
UK
Telephone +44(0)117 9656261 extension 2253
Fax +44(0)117 9763873
e-mail
melvyn.smith@uwe.ac.uk
Last up-dated: February 1999
