Camera Pose Estimation of Large Infrastructure

Camera Pose Estimation of Large Infrastructure

Problem Statement

In structural health monitoring of infrastructure such as bridges, 3D reconstruction is an essential component, and within 3D reconstruction, camera pose estimation is a critical upstream task that significantly determines the quality of 3D reconstruction. Currently, there is a lack of datasets for camera pose estimation and 3D reconstruction in large-scale infrastructure, as well as a lack of relevant research on camera pose estimation and 3D reconstruction for large-scale infrastructure in sparse viewpoint scenarios.

Tasks

1. Study knowledge and software related to camera pose estimation and 3D reconstruction.
2. Collect data (videos) of large-scale infrastructure using drones and create a camera pose estimation dataset.
3. Compare the effectiveness of existing methods (traditional methods and learning-based methods) under dense and sparse viewpoints, analyzing their strengths and weaknesses.
4. Attempt to improve the performance of existing methods for camera pose estimation in sparse viewpoint scenarios.

Work

40% Theory, 40% Simulations, 20%Writing

Contact

Qingyu Xian (q.xian@utwente.nl)