Algorithm for pose estimation based on objective function with uncertainty-weighted measuring error of feature point...

Area of Science:

• Computer Vision
• Robotics
• Geometric Measurement

Background:

• Anisotropic and non-identical gray distribution of feature points on curved surfaces presents challenges for accurate pose estimation.
• Existing algorithms often struggle with uncertainty and convergence issues in complex measurement scenarios.

Purpose of the Study:

• To propose a high-precision, uncertainty-resistant algorithm for pose estimation on curved surfaces.
• To address the limitations of existing methods in handling feature point distribution and measurement uncertainty.

Main Methods:

• A novel error objective function based on spatial collinear error is developed, incorporating uncertainty via a covariance-weighted matrix.
• The optimized generalized orthogonal iterative (GOI) algorithm is employed for robust and accurate iterative solutions.
• Redundant information is leveraged to extend the field-of-view and enhance measurement robustness.

Main Results:

• The algorithm demonstrates high accuracy, with a maximum re-projection image coordinate error of less than 0.110 pixels.
• For rocket nozzle motion, maximum static and dynamic measurement errors are superior to 0.065° and 0.128°, respectively, within a defined volume.
• Experimental results validate the approach's high accuracy and significant uncertainty attenuation capabilities.

Conclusions:

• The proposed pose estimation algorithm offers superior accuracy and robustness, particularly in the presence of measurement uncertainty.
• Its performance in simulations and practical experiments suggests strong potential for various engineering applications requiring precise motion tracking.