Advanced Inspection Techniques

  • Bruce G. Batchelor


To date, Automated Visual Inspection has concentrated almost exclusively on mass- produced objects, because the high cost of designing an inspection system has made it uneconomic to apply these techniques to any other type of artifact. However, there are several possible applications areas for industrial vision systems that have been almost totally neglected to date and which we shall address in this chapter:
  1. (i)

    Objects made in small quantities. (It has been estimated that 75% of manufactured goods are made in batches of 50, or fewer, items.)

  2. (ii)

    Complex (monolithic) objects. Examples are car engine blocks, castings, mouldings and currency notes.

  3. (iii)

    Inspecting assemblies of components (e.g. pianos, printed circuit boards, electric motors, hair-dryers, in-flight meal trays, computer keyboards, etc.).

  4. (iv)

    Mass-produced goods which are deliberately made with a high degree of individual variation, to simulate their being hand made. (There has always been a high premium paid for hand-made objects, such as wood carvings, hand- thrown pottery, embroidery, painted china, etc. While it is a relatively straightforward matter to manufacture goods with the same level of variability that hand-made goods possess1, it is not yet feasible economically to inspect them automatically. If this did come about, then “individual character”, similar to that so prized in hand-made goods, could be added to many more massproduced artifacts.)

  5. (v)

    Objects which have, by their very nature, a high degree of variability. Good examples of this group are to be found in the food manufacturing industry. (Pizzas are like fingerprints - no two look exactly alike!)

  6. (vi)

    Objects which are either flexible or consist of an assembly of jointed parts.

  7. (vii)

    Natural objects which also exhibit a high degree of variability and which have ill-defined quality criteria.



Solder Joint Print Circuit Board Data Window Inspection Technique Inspection Task 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    A small amount of randomness in stitch length, or slight deviations from a straight line, might be introduced to simulate a hand-made effect.Google Scholar
  2. 2.
    AVS, Inc., 640 Taft Street, Minneapolis, MN 55413, U.S.A.Google Scholar
  3. 3.
    The criteria for inspecting food products are often expressed as rules-for human inspectors to follow.Google Scholar
  4. 4.
    Although we have not discussed the control of a camera and lens in detail, this is straightforward in a language with the expressional power of Prolog*. Extra hardware is needed.Google Scholar
  5. 5.
    The user can employ the pull-down menus while the yesno dialogue box is visible.Google Scholar
  6. 6.
    The balloon predicate defined later does essentially the same thing.Google Scholar
  7. 7.
    In the plant dissection application, there is a similar situation where the stem is the major feature to be analysed. The same remarks apply to this type of image.Google Scholar
  8. 8.
    Mr I.P. Harris.Google Scholar
  9. 9.
    This can be defined as a VCS macro. The Autoview image processor possessed a primitive operator for performing this function.Google Scholar
  10. 10.
    In LPA MacProlog, it is necessary to define a window as being of type “Data Window”, so that the programmer can see the results of the operations assert and retract. Moreover, it is necessary to place a “seed”,in such a window, in order to ensure that assert knows where to place new clauses. Hence, if MacProlog executes assert(new_clause(qwerty,zxcvb) the new clause will only be placed in the Data Window if there already exists a clause with the same name. (The new clause will still be asserted but will not be visible to the programmer in the Data Window.) One convenient way to ensure that new clauses are visible is to place a “seed”,clause with the same name but with a different arity in the Data Window. Such a “seed”,clause might be new_clause.Google Scholar
  11. 11.
    With its sides at 0° and 90°.Google Scholar
  12. 12.
    Name five objects, other than bananas, that satisfy this test. Failed? Then I have proved the point!Google Scholar

Copyright information

© Springer-Verlag London 1991

Authors and Affiliations

  • Bruce G. Batchelor
    • 1
  1. 1.School of ElectricElectronic and Systems EngineeringCardiffUK

Personalised recommendations