Abstract: Aiming at the problem that   multi-source heterogeneous data had the complex sources and unique structure, resulting in a greater difficulty in its fusion. In order to improve the efficiency and accuracy of data fusion, we proposed a multi-source heterogeneous data fusion model based on fuzzy mathematics. Firstly, by utilizing a federated weighted average fusion strategy, the metadata transmitted from various sensors to the data level fusion layer was integrated to obtain the data level fusion results. Secondly, combined with the principal component analysis method and canonical correlation analysis method, the features of data unified by Web Ontology Language were extracted to complete the  feature level data fusion. Thirdly, a fuzzy rule library established and updated based on fuzzy mathematics theory was used to obtain decision level fusion results through decision fusion algorithms. Finally, we combined the data fusion results of above different levels to establish a data fusion model, and obtained the final data fusion result. The experimental results show that the maximum covariance value and absolute error value of the proposed method do not exceed 0.15, and the shortest fusion time is only 12.6 ms. The fusion accuracy and stability of this method are good, and both timeliness and anti-interference have significant advantages.

Key words: fuzzy mathematics, typical correlation analysis method, fuzzy rule, multi-source heterogeneous data, data fusion