As the track before detect algorithm of small targets has tracking inefficiency and complex computation problems under low signal-to-noise ratio (SNR) condition, a small targets detection and tracking algorithm is proposed by using the box particle probability hypothesis density filter. Firstly, in consideration of the images of the targets'measured intensity within targets'contribution intensity and noise, image noise is restrained using the mean filter. Then, the interval measurements are selected from the region of the larger intensity value based on the uncrossed principle. Finally, according to the interval analysis technology, the proposed interval measurements are used to track the targets using the box particle probability hypothesis density (BOX-PHD) filter. Simulation results show that the proposed method can improve the target tracking performance and computation effectively.