In this paper we use range profiles as the feature vectors for data representation, and use the matching score method and the neural network approach to automatically identify aerospace objects. The backpropagation model is employed in the neural network approach to train a neural network. Recognition performances with different training methods are compared and discussed. Experimental results show that the neural network approach has a little greater capability in identifying similar feature vectors and has stronger immunity to Gaussian noises than the matching score method does in certain cases. However, the latter method is range-shift invariant while the former method may fail to recognize a shifted feature vector. We propose a centroid-aligned method to overcome the range shift problem. Simulated results show that this method is simple and effective.