| Summary: | 碩士 === 華梵大學 === 機電工程研究所 === 89 === Abstract
The IC packaging substrates such as lead-frames and BGA are required to pass very rigorous quality assurance. Although most of the imperfection of the substrates can be detected and winnowed by using the automatic vision inspection machine, but the surface defects on the substrate can only be identified and classified artificially.
This paper raises a new idea and a performing procedure that is to apply an imaging process method to describe the feature of various surface defects quantitatively. The criterion for classifying the various defects can be set according to the value of the threshold be determined by the histogram of the accumulated data. Here we have selected seven image process methods to apply to the five surface defects respectively. The seven imaging process are erosion-dilation, degree of a circular, convolution, gradient of gray level, fast fourier transform, centroid of contrast, correlation, and the distribution change of gray level. The five surface defects are pit, foreign materials, nodule, rust and silver sludge.
The experimental results showed that the recognition yield for all the defects was increased from 80% to 97% whenever the total amount of samples was examined from 100 to 500 pieces. By choosing the imaging process methods properly, the recognition yield of the defects can be achieved to be 100%.
Keyword: Leadframe, BGA, recognition yield, surface defect, imaging process
|
|---|
