An assessment of the capability of artificial neural networks to inter
pret radar images of reinforced concrete is presented. The capability
of three-layer, fully connected networks to detect the presence of a b
ar, the size of a bar, and the depth of a bar is examined. The study d
emonstrates that the use of a neural network approach to interpret com
plex sub-surface radar results of embedded reinforcing bars is promisi
ng. The neural network approach is successful in locating reinforcing
bars over a range of embedment depths on almost all occasions. In addi
tion it is shown to be possible to place each bar into a depth categor
y with a high degree of success.