The Research Of SAM's Launching Area Calculating Based On The Deep Learning

Surface to Air missile is an important method of important places' and homeland security's defense.It played an important role in lots of wars.The launching area of surface to air missile is an important parameter for commander to decide the launching time.Launching area can be divided into theoretical and practical launching area,theoretical launching area is got from equation of missile motion;the actual launch area is the analysis and solution of the data according to the theory.The actual launch area calculation accuracy is one of the most important tactical performance indexes of the defense missile.It is also customer's greatest concern.Polynomial fitting is the most traditional method in surface to air missile launching area actual calculation methods.According to the polynomial function is difficult to determine,partition function difficult to grasp the scope of the problem,using BP neural network method is presented in this paper and the method fits the overall data and does not need the specific function form,.For the structure of BP neural network is difficult to determine,this paper proposes a combination of expert experience and user needs to determine the neural network structure of the neural network.When it faces a large number of data,and the relationship between the data is more complex,through the expansion of the depth of the neural network and cannot complete learning data features and enhance the accuracy of the fitting.This paper introduces the Belief Nets Deep(DBN)method to avoid the BP neural network's problem: it is easy to fall into the local optimum and expanding the depth of the network after the fitting accuracy is not significantly improving the problem.In the data preprocessing of deep learning method,the paper uses Gaussian normalization.This method can normalize the data effectively and minimize the reconstruction error.For the selection of the method of deep learning,The main work done in this paper is:1?By modeling and simulation of missile and target's movement,so as to obtain theoretical surface to air missile launch area envelope data,provide the sample data for the subsequent fitting work;2?Proposed using BP neural network to solve the ground to air missile launching area calculation,and puts forward the criterion of combination of expert experience and user needs of BP neural network in network structure is difficult to establish the problem.Then the simulation experiment is carried out according to the criterion;3?According to the BP neural network fitting method,there is no significant improvement in the accuracy of the network depth expansion,and even some cases of the problem of the decline of the fitting accuracy.This paper introduced deep learning approach,and on the use of the deep belief network of deep neural networks to solve surface to air missile launching area problem solver,research and analysis have been done;4?By comparing the experimental simulation,analysis of hidden layers and hidden nodes number for the fitting results,according to the needs of users to determine the most appropriate BP neural network structure,at last,through the comparison of the deep approach to learning of the depth of the neural network and BP neural network fitting results to find the fitting effect of the depth of learning methods of neural networks in depth is more excellent.