In the target tracking system, a core problem is how to accurately estimate the state of moving target based on sensor measurements. The traditional filtering estimation algorithm is just a reprocessing of measurements, which is the base of target tracking system. Measurements not only depend on the performance of the sensor, also depend on the relative position between multi-sensor and target. In this paper, an optimal placement of multi-sensor based on mutual information is deduced, which improves the precision of target tracking system. Simulations have verified our theoretical results.