NEURAL NETWORKS INSPECTION SYSTEM FOR GLASS BOTTLES PRODUCTION - A COMPARATIVE-STUDY

This paper describes a vision system that detects cracks in glass bott les production. The first step consists in collecting prototypes of bo ttles with and without defects. A sequence of 16 images is captured by a matrix camera while each bottle rotates in front of a specific ligh ting system. The second step is concerned with morphometric and photom etric features extraction. The subsequent decision step is performed b y different neural networks, such as MLP, RBF, PNN and LVQ. Finally, p erformances of these networks have been compared. All the images of bo ttles without defects have been recognized but a few images with small cracks, which are very important defects, have not been identified. H owever, since each bottle is represented by a sequence of 16 images, c racks will appear in at least three or four images, so that a defectiv e bottle can be detected at least one time through the sequence. There fore the decision system recognizes good and defective bottles with a very high rate of success.