Virtual sensors for 2D vector field tomography.

Area of Science:

• Applied Mathematics
• Image Reconstruction
• Computational Physics

Background:

• Tomography is crucial for reconstructing 2-D vector fields.
• Regular sensor placement along domain boundaries is convenient but suboptimal for accuracy.
• Uniform sampling in the Radon parameter domain yields accurate reconstructions but requires impractically many sensors.

Purpose of the Study:

• To develop a sensor placement strategy for 2-D vector field tomography that balances sensor count and reconstruction accuracy.
• To investigate the effectiveness of interpolating measurements for achieving uniform Radon domain sampling.
• To quantify the trade-off between sensor reduction and increased error using cubic spline interpolation.

Main Methods:

• Proposed uniform sensor placement along the reconstruction domain boundary.
• Utilized cubic spline interpolation to estimate measurements for uniform Radon domain sampling.
• Analyzed reconstruction errors with the proposed method versus standard approaches.
• Examined the impact of noise on reconstruction accuracy.

Main Results:

• Achieved a 60-fold reduction in sensor requirements using cubic spline interpolation.
• Observed only a ~10% increase in vector field estimation error with the proposed method.
• Ignoring uniform Radon domain sampling increased reconstruction error by ~30% compared to the proposed method.
• Evaluated the influence of noise on the tomographic reconstruction.

Conclusions:

• Uniform sensor placement combined with cubic spline interpolation offers a practical and efficient solution for 2-D vector field tomography.
• The proposed method significantly reduces hardware demands without substantial loss of accuracy.
• This approach provides a superior alternative to traditional methods that either compromise accuracy or demand excessive sensors.