One of the most important problems in automated integrated circuit (IC
) board inspection is to recognize the codes on the IC. In this paper,
we propose a method that is suitable for the recognition of the stamp
ed characters on the IC. It first uses the loss function to establish
a classifier. Then, by means of feature extraction the similarity meas
ure between the recognizing pattern and each of the 36 characters is o
btained. Finally, the pattern is determined by choosing the character
with the maximum similarity. Furthermore, a character matching method
is proposed to increase the efficiency and effectiveness of inspection
. This matching method can also reduce the false alarm rate due to poo
r printing quality.