The Study of Combined Invariants Optimization Method on Aircraft Recognition

The method, which is interested in controlling the stability of image invariant features at every stage, is proposed to extract and select new combined invariants for training classifier when aircraft types are recognized. First, a typical aircraft automatic recognition system based on images is analyzed. Second, Hu´s moments, Affine moments, Normalized Moment of Inertia and Normalized Fourier Descriptors are introduced. Third, multiple images with different kinds of 3D aircrafts under various small space angles are collected and the above four invariants from these images are extracted. Fourth, the new combined invariants are constructed based on these four kinds of invariants and are sent to support vector machine classifier for recognizing aircraft types. At last, the simulation results are shown that the recognition rate will be improved apparently if the new optimized combined invariants are used to training the support vector machine classifier.