By focusing on the multi-sensors’ measurement on the same characteristic index from different directions, a multi-sensor data fusion algorithm based on optimal fusion set is proposed. The optimal fusion set is defined to obtain the valid observation data, in which the consistent fusion degree and the distribution degree of sensors are measured by using an optimal fusion degree matrix set. Next, the weights of the valid observation data in the fusion process are reasonably assigned. The final expression of data fusion and estimation is obtained. The data fusion results show that this method demonstrates higher fusion precision, and also excellent.