An improved Bernoulli particle filter algorithm based on unscented transformation is proposed for target detection and tracking in the interval measurement. Under the theory framework of the particle filter, an algorithm which combines the particle filter with the unscented Kalman filter(UKF) is presented. When persistent particles are calculated during the predicted measure by using the algorithm, the persistent particles are most likely to be in the region of high likelihood based on the current measurement, which makes the particles distribution more approach to the true posterior distribution of the state. Simulation results show that the tracking error of the improved Bernoulli particle filter is less than the original algorithm.