Sequential Aircraft Target Recognition Algorithm Research Based On LLE And HMM

The aircraft target recognition has become one of the important areas of automatic target recognition (ATR) system, because it has very important applications in such aspect as precision-guidance, identification friend or foe, etc. With the rapid development of image processing, pattern classification and artificial intelligence, the aircraft target recognition will become a major component of the current and future weapons. But, the machine recognition for the aircraft target is hardly used in real applications, because of the complexity of the posture changes for the flying aircraft. It’s still challenging to raise the low recognition rate caused by the aircrafts’posture changes. Aiming to solve the problem above, we analyse the images for the aircrafts during flying. A further study on aircraft flight process is conducted by all postures model and sequential information. The main work is described as follows.On the basis of the research of the mainstream ATR algorithm of aircrafts, we analyse all postures model for the aircrafts. An aircraft target recognition algorithm based on locally linear embedding (LLE) and hidden markov model (HMM) is proposed. The specialty of the algorithm in this paper is:Firstly, partial singular values of outline are extracted as the image features. Secondly, locally linear embedding method is used for nonlinear dimension reduction. Thirdly, to get a set of observation sequence information, preliminary identification of Nearest Feature Line Classifier (NFL) is used in the low dimensional embedding space. At last, in order to make up for the loss of the data information in the dimension reduction, at the same time according to the similarity of image for the different aircrafts in some attitudes, sequence information fusion based on HMM is used to complete aircraft target recognition.Finally, according to the simulation experiment results on sequential aircraft target, the parameter in this algorithm was seted up, and it turns out that the algorithm in this paper can maintain high exact recognition rate in the situation of the aircrafts with greatly varied postures or even the aircrafts occulted in the aircraft sequence, which means the validity of this method is shown by the simulation experiment.