Photogrammetry and 3D computer vision aim at recovering the structure of real-world objects/scenes from images. This course is about the theories, methodologies, and techniques of 3D computer vision for the built environment. In the term of this course, students will learn the basic knowledge and algorithms in 3D computer vision through a series of lectures, reading materials, lab exercises, and group assignments. The topics cover the whole pipeline of reconstructing 3D models from images:
- Cameras models: how a point from the real world gets projected onto the image plane and how to recover the camera parameters from a set of observations;
- Epipolar geometry: the geometric relations between 3D points and their images points; the constraints between the image points;
- Image matching: define and match image features (SIFT) to establish correspondences between images;
- Structure from motion: recover/refine geometry and camera parameters from a set of images;
- Multi-view stereo and learning-based approaches for recovering dense geometry (e.g., point clouds) from images;
- Surface reconstruction: obtain 3D surface models of real-world objects from point clouds.
After finishing this course, the students will be able to:
- apply linear algebra knowledge to implement basic 3D computer vision algorithms;
- explain the main concepts in 3D computer vision (i.e., camera models, epipolar constraints, fundamental matrix, image matching, triangulation, structure from motion, bundle adjustment, multi-view stereo, and surface reconstruction);
- explain the principles of the state-of-the-art 3D computer vision pipelines for 3D dense reconstruction from images;
- evaluate methods for reconstructing smooth surfaces and piecewise planar objects, and choose applicable methods to solve specific reconstruction problems;
- propose and implement solutions for reconstructing real-world buildings from images.
