THESIS
2016
xiii, 89 pages : illustrations ; 30 cm
Abstract
Injection moulding is a typical rapid production, low-cost batch process. Plastic parts are produced
in such an enormous amount by the injection moulding that it is impossible for each item to
be inspected by controllers.. Therefore, parts are inspected randomly for quality assurance. However,
computers and cameras allow the products to be examined at high speed which makes full
inspection with stable performance possible. As the devices can function 24 hours a day, minus
some maintenance down time, they are more favorable than human inspectors with their limitations..
Motivated by this, a device based computer vision to inspect plastic parts, as well as a new
method to evaluate the quality for these parts, has been developed.
The device was designed combining the traditional vi...[
Read more ]
Injection moulding is a typical rapid production, low-cost batch process. Plastic parts are produced
in such an enormous amount by the injection moulding that it is impossible for each item to
be inspected by controllers.. Therefore, parts are inspected randomly for quality assurance. However,
computers and cameras allow the products to be examined at high speed which makes full
inspection with stable performance possible. As the devices can function 24 hours a day, minus
some maintenance down time, they are more favorable than human inspectors with their limitations..
Motivated by this, a device based computer vision to inspect plastic parts, as well as a new
method to evaluate the quality for these parts, has been developed.
The device was designed combining the traditional vision system with an additional projector with
structured lights projected on the testing object. These structured lights code all the points on the
surface of the tested objects. The features of the image are strengthened by Unique codes for every
single point. The correspondence of the pair points from the left and right images could be found
by matching the indexes of the points without having to process the images and match the image
features.
Based on the data obtained by the quality inspection device, a new type of method should be developed
in order to evaluate the quality derived from the 3D information. The Iterative Closest Point(ICP) algorithm was used to create the new quality evaluation method by computing the value
“distance” between the closest points of the tested object and the perfect reference. The Kd-tree
data structure improved the speed performance of this method.
After testing six groups of plastic products, some conclusions have been drawn. The results show
the development of a proposed device and an evaluation system that can effectively sift unqualified
products out of manufactured batches. However, for the future, a problem regarding blind zoom
needs to be solved using motion cameras, and some intelligent algorithms are required in order to
adjust the threshold of the evaluation system.
Post a Comment