To address the issues of sparse laser point cloud projection on images and the inadequacy of single sensor-based positioning accuracy, a tightly coupled visual-laser odometry with dynamic gain is proposed. First, an adaptive point cloud depth enhancement algorithm is used to improve the accuracy of feature point depth estimation. Then, a residual function is constructed based on visual feature matching and laser feature matching. Three factors that contribute to the increase in residuals are analyzed, and different dynamic gain functions are designed to compensate for the errors. Experimental results show that the proposed algorithm improves the accuracy compared to traditional visual-laser fusion localization algorithms, enhancing the global consistency of the positioning trajectory.