Yes; DA may be used with group sizes are unequal. There are two important points to keep in mind. First, the measure variances and covariances will generally be best estimated for the group having the most cases and least well estimated for the group having the fewest cases.
Second, your choice of prior probability of group membership will have an influence on the classification accuracy of the discriminant solution when groups are unequal in size. IBM SPSS defaults to equal group membership priors, I believe, though you can, via the dialog boxes, specify priors that are proportional to observed group sizes. If you use syntax, you can specify any set of priors you wish.