Due to the climate characteristics, the maintenance and management of track engineering in seasonal frozen areas face more challenges. In order to improve the information management effectiveness of track engineering in seasonal frozen areas, this study proposes a three-dimensional geographic information fusion method for track engineering in seasonal frozen areas based on a cascaded forest model. Firstly, the UAV photogrammetry technology is used to obtain the 3D geographic information data of track maintenance in the seasonal freezing area. Then, denoise the original image data. Finally, build a cascade forest model, input the preprocessed image data, carry out feature extraction and layer by layer fusion for each image data, and obtain the 3D geographic information fusion results of track maintenance in seasonal frozen areas. The experimental results show that after applying this method, the plane error of the information fusion result is only 0.22 m, and the elevation error is only 0.26 m, indicating that this method has a high accuracy in three-dimensional geographic information fusion.