Object recognition in radar images using conjugation indices and support subspaces

Abstract

We suggest an object recognition method based on synthetic aperture radar images. The so-called conjugation index between the vector under recognition and a subspace composed of vectors of a training set has been used as a distance function. The processes of clustering have been constructed using support subspaces. Different processes of the training set resampling through the exclusion of vague vectors from the set using the conjugation index have been discussed. The dependence of the recognition quality on the support subspace dimension has been analyzed. The results of experiments demonstrate that the proposed method provides a higher recognition quality than that offered by the support vector method (SVM).