Aiming at the problem of positioning and map construction in the intelligent unmanned system, a visual inertial SLAM method based on particle filter is proposed. This method is designed and implemented based on the traditional particle filter idea. In the back-end processing, only the pose state is filtered, which effectively solves the dimensional explosion problem of the traditional algorithm, reduces the amount of calculation while ensuring a certain degree of accuracy, and takes into account the accuracy and immediacy requirements of SLAM. The experimental results show that the relative error of the overall positioning accuracy of this method is less than 5%, and the effect is better in small scenes with suitable lighting conditions, which can achieve decimeter-level accuracy. It is proved that the SLAM method can fulfill the requirements of SLAM, realize the functions of simultaneous localization and mapping. The method has certain accuracy, stability and robustness. Particle filtering can be applied to the field of visual inertial SLAM and achieve higher accuracy requirements.