| Summary: | 碩士 === 中華大學 === 機械與航太工程研究所 === 89 === To shorten the inspection time and ensure quality of products, the thesis applied color machine vision technique instead of graylevel to inspect BGA substrate. Using the color difference among gold, nickel, copper, and solder mask, it will be able to detect any defect inside the gold plating area. For PBGA or Tape BGA, items to be inspected include substrate surface, gold plating area, metal trace, via hole, burr, chip out, and discoloration. The gold plating area includes bond areas and ball pad. Furthermore, the bond areas include ground rings, power rings, and bond fingers. However, the thesis focused only on the inspection of bond fingers and ball pads inside the gold plating areas.
A normal bond finger or ball pad possesses a particular color, therefore, once abnormal occurs color of bond finger or ball pad will change. For example, a scratch on the bond finger may exposure nickel underneath. Consequently, the thesis based on color machine vision technique. First, apply back propagation neural network color image segmentation technique to detect all the possible defects. Then, use feature extraction and analysis as well as back propagation neural network classification techniques to class the detected defects.
To verify the usefulness of the research results, an integrated hardware and software system has been established. The developed system is capable of detecting almost any defect on the surface of gold plating area. Furthermore, the detected defects can be classed into stain, scratch, pinhole, solder mask, or unknown five types.
|
|---|
