Aiming at solving the problem of low accuracy existing in the nondestructive testing system for eggshell, a new hybrid scheme of rough sets and support vector machine for classifier designing was proposed. First, redundant characteristic parameters were reduced based on the rough sets theory. During reducing process, the characteristic parameters were quantified by fuzzy C means clustering algorithm, and the condition attributes were reduced via heuristic search according to their own importance. And then, the classifier was trained by support vector machine based on the reduction result. During training process, the classifier model parameters were optimized by cross validation. The experiments show that the accordance rate of the proposed method can reach 94.6%, which has great engineering application perspective.