Abstract: A hierarchical multi-sensor data fusion model was proposed in order to improve precision of early warning of multi-sensor detection system and reduce rate of unknowing or false judgement of the state in monitoring process of the multi-sensor. The model uses hierarchical analysis method to determine membership degree and the corresponding weights in data level, uses fuzzy evaluation method to fuse data in feature level, and adopts D-S evidence theory to fuse data in decision level. It can calculate with different algorithms according to kind of the input fusion data. The experimental results show that the model can increase membership grade of security, slight, dangerous states 8.3%, 6%, 29.2% compared with initial data, and verify the effectiveness of the model.