研究目的
To develop a linear method for self-calibration of a projective camera by observing a deforming object represented by a trajectory basis, reducing unknowns and improving robustness.
研究成果
The proposed linear self-calibration method using trajectory basis effectively reduces unknowns and improves robustness, as demonstrated by lower errors and variances in simulation and better 3D reconstruction in real data compared to existing methods, enabling accurate camera calibration for non-rigid objects.
研究不足
The method assumes trajectories span a low-dimensional subspace and relies on predefined DCT basis, which may not capture all deformation complexities; noise sensitivity increases with deviation, and it is specific to perspective projection cameras.
1：Experimental Design and Method Selection:
The method involves modeling deforming object trajectories using predefined DCT basis vectors to reduce unknowns and formulate camera self-calibration as a linear optimization problem.
2：Sample Selection and Data Sources:
Simulation data with 100 spatial points and real data from dinosaur and dance image sequences are used, with feature points extracted and tracked.
3：List of Experimental Equipment and Materials:
A personal computer with Matlab
4：0, and image sequences from simulated and real scenarios. Experimental Procedures and Operational Workflow:
Depth factors are solved based on low-rank matrix properties, SVD decomposition is applied, absolute conic is derived from linear equations, and camera intrinsic parameters are obtained via Cholesky decomposition.
5：Data Analysis Methods:
Performance is evaluated by comparing mean and variance of camera parameters with added noise, and 3D reconstruction results are visually assessed.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
