Abstract: In order to solve the problem of the single piece of sensors' measurement and the complex and repeated measured data in the coal mine dust monitoring, a new method that taking the dust concentration and dust particle size as the monitoring objects is proposed by using the data fusion technology to process the data of sensor. On the basis of the establishment of data fusion two-stage structure model, we first applied fusion algorithm based on matrix analysis to carry out data level information fusion of homogeneous data source, and then D-S evidence theory was applied to heterogeneous source data for decision level fusion. Finally, we realize the integration optimization of sensor information. Test and application results show that this method is perfect in dust characterization evaluation, which significantly improves the accuracy and reliability of the sensor information.