Visual Inertial Odometry (VIO) is commonly utilized for the localization of mobile platforms such as autonomous vehicles and drones. However, long-time movement can cause VIO drift. Although it is an effective method of combining prior map information to provide a more accurate localization result, it is usually costly to build a global and consistent map for large scenes. In an effort to mitigate the accumulation of error intrinsic to long-duration deployment of odometry and to avoid the construction of a global map, our approach involves the segmentation of the scene or trajectory for discrete odometry operations. We introduce a method for triggered reinitialization of the VIO, which incorporates tightly coupled map information derived from multi-segment motion processes into the VIO system. This strategy effectively reduces the accumulation of VIO drifts, a common challenge in robotics.