Research On Sequential Aircraft Target Recognition Based On ELM And HMM

With the rapid development of image processing,machine learning and artificial intelligence,automatic target recognition system plays an increasingly important role in the development of the present and future weapons.Among them,aircraft target recognition,as an important part of the ATR system,has a very important role in the automatic monitoring and detection,precision guidance and the identification of the enemy and so on.In the process of aircraft target identification,due to the changeable attitude of aircraft,a wide range of external environment interference and other reasons,target aircraft recognition rate is relatively low.In this thesis,we carry out research on the aircraft flight process,utilize the multi-feature information fusion and sequence information fusion,establish a full attitude model for the aircraft.To a certain extent,the algorithm solve the problem of low recognition rate caused by changeable pose,in the meanwhile,taking the algorithm’s real-time performance into account.The main work of this paper is as follows:Based on the study of the target identification algorithm of mainstream aircraft,Hidden Markov Model,which is commonly used in pattern recognition,is introduced to model the full pose of the flight process,and the observation value is constructed by multi-feature fusion method.A sequential aircraft target recognition algorithm based on ELM and HMM is proposed.The main features of this algorithm are as follows:Firstly,the local singular value of the aircraft target is extracted and the scale-invariant feature transform is used as the image feature;Secondly,the ELM algorithm is used to train the single hidden layer feedforward neural networks,which is used as classifier to establish Basic Belief Assignment.Then using the PCR5 formula of DSmT fusion theory to implement multi-feature fusion of two different features of the image.Thirdly,the HMM is modeled for each type of aircraft,it is one of the major innovations of this paper to describe the whole attitude flight process of the aircraft by using the 7 middle attitude as the state value of the HMM.Finally,according to the obtained target observation sequence,the similarity between the observation sequence and the various aircraft models is calculated to complete the identification of the aircraft target.Finally,the simulation results show that the average recognition rate of the aircraft can reach 95.4%,and it can still maintain a high recognition rate under the conditions of frame occlusion and scaling of the aircraft sequence.In real time,the algorithm also has certain advantages.