基于模糊神经网络的多源信息融合

发布时间：2018-01-21 11:28

本文关键词： 信息融合模糊神经网络主成份分析雷达回波滤波算法　出处：《中北大学》2017年硕士论文　论文类型：学位论文

【摘要】：多源信息融合也称多传感器信息融合,是把多个渠道、多个方位采集的局部环境的不完整信息加以综合,消除多源信息间可能存在的冗余和矛盾的信息,加以互补,降低其不确定性,以形成对系统环境的相对完整一致性描述的过程,从而提高智能系统的决策、规划、反映的快速性和正确性,降低决策风险。本文首先利用模糊神经网络进行多传感器信息融合,与传统的以概率理论为基础的融合方法相比,无需任何先验信息;克服了证据理论融合方法中的证据难以获得、计算量大的缺陷。不仅拓宽了神经网络处理信息的能力,使其既能处理精确信息,也能处理不精确信息或模糊信息,它将模糊推理机制蕴涵于神经网络结构之中;而且,训练好的模糊神经网络无需额外信息,就可融合多传感器信息,解决了模糊理论融合多传感器信息中的模糊规则的提取和隶属函数的确定等,提高了融合系统的融合能力以及融合的准确性。其次在T-S模糊神经网络数据融合的基础上,基于标准T-S模糊融合算法的时间和空间复杂度进行算法和网络结构的改进。通过仿真实验验证了改进的算法在融合过程中的合理性、稳定性和准确性。并且通过Kohonen网络的聚类算法来选取适当的隶属度函数个数,通过实验验证了聚类算法在T-S模糊神经网络数据融合算法中运用的合理性和有效性。然后以模糊神经网络为基础对雷达目标回波的识别进行研究和实验,引入两种比较实用的滤波算法对雷达目标回波的数据进行优化。通过仿真实验证明,滤波的优化效果十分显著。最后,采用主成份分析算法对用实地采录的雷达目标回波数据进行特征提取。实验结果证明,本文采用的方法在雷达目标回波的识别上具有很高的准确率,并具有良好的泛化性能。
[Abstract]:Multi-source information fusion, also called multi-sensor information fusion, integrates incomplete information of local environment collected from multiple channels and azimuths to eliminate the possible redundancy and contradictory information between multi-source information. It can complement each other and reduce its uncertainty so as to form a relatively complete and consistent description of the system environment, thus improving the speed and correctness of the intelligent system's decision-making, planning and reflecting. Firstly, the fuzzy neural network is used for multi-sensor information fusion. Compared with the traditional fusion method based on probability theory, no prior information is needed. It overcomes the defect that the evidence in evidence theory fusion method is difficult to obtain and has a large amount of calculation. It not only broadens the ability of neural network to deal with information, but also makes it capable of dealing with accurate information. It can also deal with imprecise information or fuzzy information, which implicates the fuzzy reasoning mechanism in the neural network structure. Moreover, the trained fuzzy neural network can fuse the multi-sensor information without additional information, which solves the problem of fuzzy rule extraction and membership function determination in fuzzy theory fusion multi-sensor information. The fusion ability and accuracy of fusion system are improved. Secondly, based on T-S fuzzy neural network data fusion. The algorithm and network structure are improved based on the time and space complexity of the standard T-S fuzzy fusion algorithm, and the rationality of the improved algorithm in the fusion process is verified by simulation experiments. Stability and accuracy. And the Kohonen network clustering algorithm to select the appropriate number of membership functions. The validity and rationality of the clustering algorithm in T-S fuzzy neural network data fusion algorithm are verified by experiments. Then the recognition of radar target echo is studied and tested based on fuzzy neural network. Two practical filtering algorithms are introduced to optimize the radar target echo data. The simulation results show that the optimization effect of the filter is very significant. Finally. The principal component analysis (PCA) algorithm is used to extract the feature of radar target echo data recorded in the field. The experimental results show that the method adopted in this paper has a high accuracy in radar target echo recognition. And has good generalization performance.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212;TP183

【参考文献】