Abstract: In view of problems of inaccurate mine safety monitoring data and unreliable safety evaluation, a two-level data fusion method based on multi-sensor data fusion was proposed. The first level fusion uses batch estimation algorithm to process single sensor data in order to improve accuracy of data collection, then adopts adaptive weighting algorithm to process similar multi-sensor data in order to obtain fused value of mine environmental parameters. The second level fusion makes grey correlation degree analysis between the fused value of mine environmental parameters and mine safety standard feature vector, so as to get consistent estimation of mine safety status. The example verifies feasibility and effectiveness of the method.