The invention discloses a video target tracking method based on improved box particle filtering, and mainly solves the problems of poor search and capture performance and low particle sampling efficiency in large dynamic range target tracking carried out by a traditional particle filtering video target tracking method. The video target tracking method comprises the following implementation steps: 1) under a particle filtering frame, expanding common point particles into box particles, obtaining a candidate target through target state prediction, and extracting the characteristics of the candidate target; 2) calculating a distance between the characteristics of the candidate target and a template, contracting the box particles, and obtaining corresponding weights; and 3) resampling the box particles so as to carry out fusion to obtain the state estimation of a target to realize the reliable tracking of the target. The video target tracking method can efficiently cover a target state space, improves sampling efficiency, improves search and capture performance for a large dynamic range target, and can be used for video surveillance and target tracking systems.