Data Fusion, Neural Network-based Target Identification Method

With the need of modern war and the development of science and technology, data fusion as a newly arisen subject is given wide attention in recent years. Target recognition is an important part of data fusion technology, and it is also a key subject in the field of military technological research.After a general comment on the methods of target recognition, target recognition based on neural networks, this essay mainly focused on the problem of target recognition of neural network in data fusion. Under the presumption of the know measure vector, neural network is applied to target recognition. Firstly, the essay puts forward the method of using radial basis function neural network to recognize multi-targets. Concerning the shortcomings of the traditional radial basis function neural network model algorithm, the essay brings forth improved learning algorithm and employs the new algorithm in the target recognition. Through simulation the essay compares the advantages and disadvantages of the improved and traditional radial basis function neural network algorithm. Then the essay studies the target recognition of self-organizing neural network. Due to the existing shortcomings of self-organizing neural network, the essay puts forward improved methods, and through simulation, the availability of new algorithm has been proved. As to the new algorithm's high wrong recognition rate in target samples overlapped, it has been combined with the LVQ net and used to target recognition, and the simulations indicate that the finally improved algorithm increases the target recognition rate. Finally, the essay brings forth a method of neural network target recognition based on genetic algorithm. Because of the disadvantages of traditional genetic algorithm, the essay made some improvement of it. Through two test functions, the essay compares the iterative times of the improved and traditional genetic algorithm. At the same time, in order to overcome the certain blindness and randomicity existing in the process of the best-method searching of the traditional genetic algorithm, the essay combines the method with BP supervised algorithm forming a new algorithm. Furthermore, the combined algorithm is applied to target recognition. The simulation suggests that the new algorithm has a higher rate than the BP algorithm. The essay put forward the further research of target recognition based on neural network in the end.