The goal of this project is to generate accurate 3D models from aerial views for visualization and mission planning. The framework's input is the recorded raw sensor data (GPS, images, IMU) as well as camera intrinsics and extrinsics that have been estimated online during the UAV mission. First, the framework to be implemented, needs to check the quality of the input data. Based on this input data, the framework selects the subset of images (key-framing) that is needed to accurately generate a model of the terrain. If required,
variables that affect the model's accuracy (camera extrinsics, intrinsics, distortion model) need to be re-optimized using bundle adjustment. As end result, a 3d model in form of a textured mesh is generated. The pipeline is to be integrated into QT for visualization and debugging. Finally, the framework is to output an orthomosaic, digital surface map (DSM), mesh, and statistics about the quality of the dataset. For the generation of the orthomosaic and DSM the student can rely on existing code.
The goal of this project is to generate accurate 3D models from aerial views for visualization and mission planning. The framework's input is the recorded raw sensor data (GPS, images, IMU) as well as camera intrinsics and extrinsics that have been estimated online during the UAV mission. First, the framework to be implemented, needs to check the quality of the input data. Based on this input data, the framework selects the subset of images (key-framing) that is needed to accurately generate a model of the terrain. If required, variables that affect the model's accuracy (camera extrinsics, intrinsics, distortion model) need to be re-optimized using bundle adjustment. As end result, a 3d model in form of a textured mesh is generated. The pipeline is to be integrated into QT for visualization and debugging. Finally, the framework is to output an orthomosaic, digital surface map (DSM), mesh, and statistics about the quality of the dataset. For the generation of the orthomosaic and DSM the student can rely on existing code.
- Literature review on state-of-the-art of bundle adjustment and 3D model generation from aerial views.
- Become familiar with existing tools and datasets.
- Implement the framework as outlined above.
- Evaluate your framework on many existing datasets, recorded by various fixed-wing and rotary-wing UAVs.
- Literature review on state-of-the-art of bundle adjustment and 3D model generation from aerial views. - Become familiar with existing tools and datasets. - Implement the framework as outlined above. - Evaluate your framework on many existing datasets, recorded by various fixed-wing and rotary-wing UAVs.
- Highly motivated and independent student.
- Interest in topics related to 3D modeling, geometry, computer vision, bundle adjustment
- Course(s) related to computer vision required
- Experience in C++ mandatory
- Experience in QT is a plus
- Enrolled at ETH Zurich
- Highly motivated and independent student. - Interest in topics related to 3D modeling, geometry, computer vision, bundle adjustment - Course(s) related to computer vision required - Experience in C++ mandatory - Experience in QT is a plus - Enrolled at ETH Zurich
If you are interested in the project, please send your transcript of records, CV, and a few words about your experience in programming and photogrammetry to hitimo@ethz.ch
If you are interested in the project, please send your transcript of records, CV, and a few words about your experience in programming and photogrammetry to hitimo@ethz.ch