Aircraft Target Recognition Research Based On Multi-classifier Fusion Under Multi-views

With the development of science and technology, Aircraft target recognition is become one of the hotspots in the research of computer vision field. The plane has a broad application in both military and civilian fields. In the field of military, aircraft target recognition can be applied in Navigation guidance system, Defense system,military target detection. In the field of civilian, aircraft target recognition can provide real-time dynamic monitoring for the civil aviation airport. Through these, the airport staff can effective management the airport.The technology of aircraft target recognition has a great progress in recent years,but there are still many shortcomings of existing algorithms, such as low recognition rate of aircraft and time-consuming under multiple viewpoints and multiple positions.Therefore, aiming to improve the recognition accuracy of the algorithms, reduce the time-consuming, this paper proposes a new aircraft recognition method basis on the existing algorithms. The main content of this paper as follows:1.This thesis summarizes generalized flowsheet and the key technology of aircraft target recognition, then analyzes key technology in each step of process. In this part,which focuses on analyzing feature extraction and recognition method. Firstly,analyzes the advantages and disadvantages of characteristics in feature extraction process. Then,explores the connection between the different types of features. Finally, the various types of classifier were introduced in detail which in recognition algorithms.2.Aiming at the low recognition rate of aircraft under multiple viewpoints and multiple positions, this paper proposes an aircraft recognition method based on SVM(Support Vector Machine) classifier and DSm T(Dezert-Smarandache theory) on multiple features under multi-views. Firstly, structure multiple SVM classifier, then to complete the aircraft target recognition based on DSm T combination rules on multiple SVM classifier. The algorithm can overcome the problem of the low recognition rate of aircraft under multiple viewpoints and multiple positions. According to the experiment results, this algorithm has a high recognition rate of different types of aircraft even though aircraft under multiple viewpoints and multiple positions.3.In order to improve the recognition rate of aircraft and reduce time-consuming of the algorithms under multiple viewpoints and multiple positions. This paper proposes an aircraft recognition method based on extreme learning machine on multiple featuresunder multi-views. Firstly, extracts wavelet moment, Zernike moment and Fourier descriptor. Then design each modular of extreme learning machine based on wavelet moment, Zernike moment and Fourier descriptor. Finally, it recognizes the candidate targets based on weighted voting fusion on results of different modular of extreme learning. Based on the characteristics of extreme learning machine, the new algorithm reduce the consumption time in the flowsheet of aircraft target recognition and improve the real-time performance of the algorithm. According to the experiment results, this algorithm has a high recognition rate of different types of aircraft and less time-consuming even though aircraft under multiple viewpoints and multiple positions.