Automatic target recognition of ISAR object images based on neural network

In this paper an integrate approach of automatic target recognition is proposed with a three-feedforward neural network. It includes images pre-processing, feature extraction and automatic target recognition and classification of ISAR object images. Using the log-spiral transform after moving the centroid of ISAR images to the original point makes the target recognition invariant on translation, rotation and scale. The approach improves the ratio of accurate recognition and reduces the amount of calculation. The results of experiments with field data show that the approach is effective.