In this paper, we focus on using the sensor network to track a moving target. We estimate the trajectories of moving targets by collecting the information from sensors measurements, and assume that sensors can be randomly moving within a limited radius r. Also assuming each sensor is able to calculate/obtain the distance between itself and the moving target. Here we introduce the modified particle filter (MPF) algorithm. MPF means PF with varying particle numbers. For a nominal PF algorithm, particle number is fixed and we define the nominal one to be the so called fixed particle filter (FPF). Simulations show MPF sustains a smaller estimation error and at the same time substantially reduce the computational load.