A Robust O(n) Solution to the Perspective-n-Point Problem.

Area of Science:

• Computer Vision
• Robotics
• Geometric Computing

Background:

• The Perspective-n-Point (PnP) problem is fundamental in computer vision for determining object pose from 2D-3D correspondences.
• Existing iterative solutions can be computationally expensive and may converge to local optima.

Purpose of the Study:

• To develop a robust and efficient noniterative solution for the PnP problem.
• To achieve high accuracy comparable to or exceeding iterative methods, particularly in challenging scenarios.

Main Methods:

• A novel approach involving dividing reference points into 3-point subsets to generate fourth-order polynomials.
• Constructing a cost function by summing the squares of these polynomials.
• Finding the optimal solution by determining the roots of the derivative of the cost function, leading to a seventh-order polynomial.

Main Results:

• The proposed noniterative method demonstrates robust performance across planar, ordinary 3D, and quasi-singular cases.
• Achieves accuracy comparable to state-of-the-art iterative algorithms but with significantly reduced computational time.
• Outperforms iterative methods in scenarios with minimal reference points (n ≤ 5).
• Exhibits efficient handling of large point sets with a computational complexity of O(n).

Conclusions:

• The noniterative PnP solution provides a computationally efficient and accurate alternative to existing methods.
• Offers superior performance in data-scarce situations, advancing pose estimation capabilities.
• The O(n) complexity makes it suitable for real-time applications and large-scale datasets.