Application of Artificial Neural Network on Multi-Sensor Information Fusion

摘要

This paper compares the Artificial Neural Networks with the previous algorithms applied to the multi-sensor information fusion of the water content detector for oil product through strictly controlling experiments and carefully processing the sample data. Numerous experiment data shows that the two types of the ANNs, BP NN and RBF NN, which are adept in the function approximation problems, can definitely improve the effect of the information fusion, but not as optimistic as estimated before. At last, the author point out that except for the structure and algorithms of the ANNs, the pre-processing procedure is also very important in the training of the NN. And the validation and integrity of the data will affect the final performance of the NN. Be careful to process the sample data and select the training set to guarantee the robust and generalization of the ANN.