Information Fusion Of Sensor Based On RBF Network

Information fusion is a kind of newly born technology; it supplies method of advancement and reliability for resolving problem of information processing and decision-making during information times. With times of internet and information coming, information fusion has become a concerned problem for national defense, industrial engineering, economy, social policy. It is no time to delay that research for theory and application of information fusion is very important. The thesis researches information fusion of sensor based on RBF network algorithm, and compare with BP network algorithm, multinomial fit that combine with Newton's interpolation method.This text mainly analyze three temperature compensation algorithms, including BP neural network, RBF neural network and multinomial fit that combine with Newton's interpolation method, through software simulation proceed the temperature compensation of measured pressure sensor data, RBF network algorithm is paid more attention. In the thesis BP network algorithm is introduced, and adopt gradient descend algorithm, applying into pressure sensor, to achieve data fitting. Multinomial fit that combine with Newton's interpolation method is introduced and applied into pressure sensor, achieve pressure value fitting by MATLAB language. By studying radial basis function neural network, a network model is constructed by two inputs and single output. High precision temperature compensation of pressure sensor is achieved by gradient descend algorithm with a momentum factor in this network model. The result is that the output of pressure sensor affected by non-objection parameters, such as environment temperatures, voltage fluctuation and so on, is overcame, and the stability and liability of the pressure sensors are improved.The results show that BP network algorithm has good characteristic of function approximation, but slow rate of constringency. Multinomial fit that combine with Newton's interpolation method has fast rate of option, but lower precision. RBF network algorithm has better characteristic of function approximation, and it is faster than BP network. The precision of RBF network is much higher than multinomial fit that combine with Newton's interpolation method.