A novel probability hypothesis density(PHD) filter for tracking multiple extended targets is proposed by using interval analysis resulting from the Gaussian mixture PHD(GM-PHD) and the recently emerged box particle filtering. The key idea is replacing traditional multiple measurements with a rectangular region of the non-zero volume in the state space, which can reduce the requirement of the measurements’ distribution. Simulation results show that, using interval analysis can reach the same tracking level of GM-PHD with a low computational complexity and a good performance on estimating the number of the targets and anti-clutter. This approach can also solve the problem of leak detection with the wrong sub-partition.