Abstract: Visual SLAM is an important navigation technology nowadays, but in practical applications, it often faces many constraints, and image shaking is a major bottleneck. This paper proposes an optimized algorithm for shaking suppression based on the combination of ORB feature point tracking and Kalman filter shaking amplitude processing. It uses a method of matching feature points in blocks to calculate the shaking amplitude, performs Kalman filtering fitting on the shaking amplitude, calculates the shaking compensation value, and compensates it to achieve high-precision SLAM mapping with shaking removal. Conduct experiments to compare the running results of improved algorithms with those of the original algorithm, using ORB-SLAM as a reference. Under slight shaking, the average error and root mean square error of the improved algorithm can be reduced by up to 7.68% and 11.19%, respectively, compared with the original algorithm. Under severe shaking, the average error and root mean square error can be reduced by up to 74.89% and 56.93%, respectively, proving that the improved algorithm has good anti-shaking effect for different scenarios and improves the accuracy of visual SLAM mapping.