ortho_mapper

A Multi-Sensor Fusion Approach for Rapid Orthoimage Generation in Large-Scale UAV Mapping

The repository contains the source code for ortho_mapper module of our system. This partial release aims to provide insight into the underlying implementation of our approach and may be useful for researchers working in related areas.

This is a simple and easy-to-use package for checking the quality of SfM through generating orthoimage quickly.

Prerequisited

• Eigen3
• OpenCV3
• PCL
• TIFF
• JSONCPP

Example

To test this package, you can use images which has EXIF for GPS, use 'dev_3.10' in colmap.

./work/shell/sfm_cam_gps_auto_calib.sh

Set proj_path in 'config/cfg.json' to your COLMAP project directory. txt_relative_path, img_relative_path, and camera_model_path are resolved relative to it; absolute paths are also accepted. A relative proj_path is resolved from the working directory. Legacy txt_path and img_path are still accepted when their corresponding relative-path fields are absent.

Image width, height, and intrinsics are read automatically from camera_model_path (default: calib/sparse/valid/cameras.txt); img_size and intrinsics are no longer needed. Supported models are SIMPLE_PINHOLE, PINHOLE, SIMPLE_RADIAL, RADIAL, OPENCV, and FULL_OPENCV. The current mapper requires exactly one camera, matching the image camera IDs. Only the intrinsic matrix is used; distortion correction is not applied.

mkdir build && cd build
cmake ..
make -j
./build/ortho_mapper config/cfg.json

The program writes result.tif and an aligned single-band, 32-bit floating point dsm.tif. DSM pixels use the same terrain samples and valid footprint as the orthoimage. Invalid DSM pixels are stored as NaN; valid values are the absolute altitudes returned by the repository's ENU-to-geodetic conversion. It also writes an aligned dsm_vis.png, using blue for low elevations, red for high elevations, and black for invalid pixels.

Citation

@inproceedings{he2025multi,
  title={A multi-sensor fusion approach for rapid orthoimage generation in large-scale uav mapping},
  author={He, Jialei and Zhan, Zhihao and Tu, Zhituo and Zhu, Xiang and Yuan, Jie},
  booktitle={2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={6808--6815},
  year={2025},
  organization={IEEE}
}