Abstract: For improving the accuracy and speed of measurement, it needs to deal with the data from multisensors. Firstly, this paper introduces the characters and training mode of BRF network, and then describes the whole processes, which includes data collection, data normalization, neural network training and network structure selection, using multisensors to detect the grain moisture based on BRF network. Lastly, trained by the net models in Matlab, the results prove that the fitted value fluctuates around target values and the differentials are less than 7%. This method has advantages and can be applied in other industry fields.

Key words: data fusion, BRF network, moisture detection