Automated Visual Inspection of Plastic Products


  • Sven Loncaric
  • Mladen Sercer
  • Igor Catic
  • Tomislav Petković
  • Josip Krapac



The main task of machine vision systems is providing computer understandable descriptions of objects from either single image or whole array of images. One of such problems is encountered in automated visual product inspection. An automated visual inspection system must discover and classify possible defects from product images and should be fairly quick and robust. These requirements are needed if automated systems are to replace human inspection which has many drawbacks, mainly caused by tiredness and slowness (modern production methods usually facilitate productions speeds humans can't cope with). The main aim of this research project is developement of algorithms for defect detection and classification along with the development of the complete prototype system.

Short description

A simple and crude laboratory prototype system is currently implemented and is used for optimization and testing of illumination. Protoype can use either background illumination or ring-shaped illumination. All images are acquired by using the protoype system, and the obtained images are used for testing of the algorithms for defect detection and classification.

Figure 1. Simple laboratory prototype system.

Curently, algorithms for detetcion of defects for various plastic products are being developed. In the images below some of the typical products are shown. In the first two images the lid made for the Franck company are shown. In the left image lid without any defect is shown, and in the right image the lid with the multiple defects is shown (both defects are shape defects). Curently a simple scheme involving Fourier descriptors and image registration is used for detection and classifiction of shape defects. More interesting problems are detection and clasification of defects for some more complicated products (like the one shown below the lids) where multiple defects can arise and the more complex shape of the product presents some difficulties.



Figure 2. a) Image of a lid with detected border drawn, b) Image of a defect lid with detected border drawn
Figure 3. Simple planar product with multiple holes and several defects.

During the development of the prototype system optimal illumination schemes were considered as the most important element of any machine vision system is image acquisition. For implementation of such systems one must first consider all possible means of image acquisition. With right illumination scheme chosen properties of an object can be made clearly distinuishable and thus significantly simplify further image analysis. In the image below optimization of parameters for the ring-shaped illumination is shown.

Figure 4. Optimization of the ring-shaped illumination.

Current work includes testing of the developed algorithms on the larger specimens of products and their further developement.