The aim was to detect boundary defects such as open, short, mouse bite and
spur on ball grid array (BGA) substrate conduct paths using machine vision.
The 2-D boundaries of BGA substrate conduct paths are initially represente
d by the 1-D tangent curve. The tangent angles were evaluated from the eige
nvector of a covariance matrix constructed by the boundary coordinates over
a small boundary segment. Since defective regions of boundaries result in
irregular tangent variations, the wavelet transform was used to decompose t
he 1-D tangent curve and capture the irregular angle variations. A boundary
defect can then be easily located by evaluating the wavelet coefficients o
f the 1-D tangent curve in its high-pass decomposition. The proposed method
is invariant with respect to the rotation of the BGA substrates and does n
ot require prestored templates for matching. Real BGA substrates with vario
us boundary defects were used as test samples to evaluate the performance o
f the proposed method. Experimental results show that the proposed method a
chieves 100% correct identification for BGA substrate boundary defects by s
electing appropriate wavelet basis and decomposition level.