摘要
针对目标识别中常用BP—DS信息融合方法识别率低,运行速度慢,抗噪性差等问题,提出一种基于PNN网络和DS证据的信息融合方法。该方法不仅综合了证据理论在处理不确定信息方面的优点和神经网络在数值逼近上的长处,利用神经网络和证据推理算法获取了基本概率赋值,同时突出了PNN网络在处理多传感器信息的准确性和运算速度上都要优越于BP网络的特点。
To solve those problems of the low recognition rate, the slow running speed and the noise immunity of the object identification method based on BP-DS model, an information fusion method based on PNN (Probabilistic Neural Network) and DS evidence theory was proposed. The method synthesized the merit of evidence theory in dealing with uncertain information and the virtue of neural network in numerical approximation, obtained the basic probability assignment function based on neural network and evidence theory, and stood out the characteristics of PNN, such as accuracy and running speed, in dealing with the information from multisensor, which were better than those of BP neural network.
出处
《海军航空工程学院学报》
2008年第3期252-256,共5页
Journal of Naval Aeronautical and Astronautical University
基金
西北工业大学本科毕业设计重点扶持项目(2007)
关键词
目标识别
多传感器
信息融合
DS证据理论
object identification
multisensor
information fusion
DS evidence theory