Abstract: In this paper artificial neuron network(ANN) technique is applied to automatic recognition of features in reverse engineering, and a kind of automatic feature recognizer is designed with good extendibility based on BP network. According to input signals derived from triangular mesh models, the feature recognizer can automatically recognize the geometric features such as generic block, boss, pocket, hole, step, slot, T-shape groove, swallow-tailed groove and V-shape groove. The proposed feature recognizer shows its characteristics of swift response and good adaptability.