The more advanced use of the automated x-ray inspection systems is a major reason to expand the concept of ‘design for producibility and testability’. In this thesis two basic cases taken from electronics production are presented which prove that the lack of the afore mentioned concept can jeopardize the inspection process itself. The caseworks presented show fiducial marks to be shielded and/or inhomogeneous due to design or production errors. The production plants dealing with those failures are using x-ray inspection systems produced by MatriX Technologies which’s standard algorithm-library contains no fiducial algorithm that could properly identify the center of the defective fiducials.
In this thesis fiducial detection methods used by other manufacturers are presented, and it is experimentally proved that they are also not capable of properly identifying the center of fiducial mark. For this reason I aimed to develop a new method.
The new algorithm consists of more and more complicated steps than its predecessor, such as morphology or segment filtering. In exchange it can shape-independently detect all the hereby presented defective fiducials with high stability.
First, the new algorithm was tested on an image database built from images made of non-recognized fiducials. All 100% of the defective fiducials were detected and it was proven that failure-free fiducials can also be stably detected by the novel algorithm.
Next, tests on many thousand boards were conducted during production in three distinct production plants in order to prove the superiority of the new algorithm over the old one. With the use of statistical hypothesis tests it could be stated by at least 95% confidence level (but in many cases even by a 99% level) that the newly developed fiducial-detecting algorithm is capable of finding the center of the fiducials more times and/or more precisely than its predecessor or any of the presented methods of substitution.