The traditional monocular SLAM system is easy to cause low localization accuracy in a weak texture environment. To solve this problem, a visual inertial SLAM system based on wire feature fusion IMU is proposed. First of all, the extraction of line features is added on the basis of the extraction of the previous endpoint features. If enough point features are not extracted in the weak texture environment, the positioning accuracy of the SLAM system will be greatly deviated, and even the tracking trajectory will be lost. Secondly, the midpoint operation of the extracted line features is taken to fuse them with the point features, and the quadtree method is used to make them evenly distributed in the image to avoid the problem of feature stacking. Finally, the visual information obtained by the monocular camera and the information obtained by the IMU were fused to achieve more accurate positioning, and the improved SLAM system was put on the Euroc data set for experiments. The improved SLAM system was more accurate and robust than the original system.