Journal of Applied Sciences1812-56541812-5662Asian Network for Scientific Information10.3923/jas.2008.2814.2824ChiuMin-Chie YehLong-Jyi HsuChe-Jung 122008816To improve the precision in the recognition process,
a new algorithm (the image division method-IDM) is proposed. Recently,
to meet various client requisitions, a flexible manufacturing process
applied to various products of smaller quantity has become the trend.
Even though the above new method can improve deficiency recognition in
PCBAs, a huge quantity of samples used in off-line training is still obligatory.
Unfortunately, the method is not suitable for a process that includes
various products of smaller quantity. Moreover, not all deficiencies can
be fully recognized by a single algorithm. To overcome the above drawbacks
and increase the recognition rate, a combination of these algorithms in
conjunction with a neural network system, which will increase the recognition
rate with fewer samples, is proposed. Consequently, results reveal that
deficiency recognition can be improved when the IDM in conjunction with
other AOI algorithms are linked with a neural network.]]>Hebb, D.O.,19491st Edn.,Hopfield, J.J.,19827925542558McCulloch, W.S. and W. Pitts,19435115133Minsky, M.L. and S.A. Papert,19691st Edn.,Yeh, C. and D.B. Perng,20042511971204Rosenblatt, F.,195865386408Rumelhart, D.E., J.L. McMclelland and C. Asanuma,19861st Edn.,