Pattern recognition techniques for hybrid microcircuits

The paper discusses pattern recognition techniques for automatic hybrid IC assembling systems. The problems concerning recognition systems are flexibility and high recognition rates for various object patterns comprised of hybrid ICs. To overcome these problems we developed two new recognition algorithms based on pattern matching methods. One method recognizes speckle pattern (printed thick-film conductor patterns) locations using matrix filtering. The other method increases the matching SNR to 4.5 times greater than previous methods using peak sharpening of the matching degree curve. These algorithms enable recognition of more than 99% (99.8% on average) of all semiconductor chip patterns (ICs, transistors, and diodes) and printed thick-film conductor patterns.