To address the issue of errors caused by external factors in different environments when applying localization algorithms with a single sensor, we propose a factor-graph-based multi-source information fusion localization method, which incorporates a multi-sensor fusion scheme and effective optimization methods. Initially, pose estimations for each sensor are obtained through feature extraction and matching. Subsequently, by constructing a factor graph model, the relationships and constraints between variables are efficiently expressed, and a dynamic weight mechanism for the optimization function is designed based on the front-end pose prediction results, allowing adaptive adjustment of the weights of each factor in the factor graph. Finally, the effectiveness of the dynamic weighting mechanism is verified on public datasets. Specifically, our method reduces position error on the KITTI dataset compared to ALOAM, LIO-SAM, and VINS-MONO by 29.35%, 32.025%, and 86.26%. On the M2DGR dataset, it reduces position error compared to ALOAM, LIO-SAM, and VINS-MONO by 61.17%, 24.975%, and 87.34%. Additionally, it enhances environmental perception capabilities and exhibits good robustness.