The precise measurement and control of the sintering mixture moisture plays an important role in improving the permeability of the mixture and improving the production and quality of sinter. Aiming at the problem that the existing moisture detection methods are difficult to accurately detect the sintering mixture moisture, an intelligent method is introduced into the moisture detection of the sintering mixture, and an intelligent online detection method of the sintering mixture moisture based on the fusion of multi-source information is proposed. Firstly, Gaussian filtering is used for image pre-processing, and XGBoost with multi-class heterogeneous image features as input is used to build a sub-model for sintering mixture moisture detection; further, the process data is pre-processed by quadrature method, and GRU network with sintering process variables as input is used to build a sub-model for sintering mixture moisture detection; the two sub-models are fused according to the features of different modal data. The results show that the moisture detection results obtained from this fusion model are more stable than that of the single sub-model and have good application value.