Research on the Optimization Method of Visual Sensor Calibration Combining Convex Lens Imaging with the Bionic...

Area of Science:

• Computer Vision
• Optimization Algorithms
• Optical Engineering

Background:

• Accurate camera calibration is crucial for various computer vision applications.
• Existing optimization methods often face challenges with local optima and limited accuracy.

Purpose of the Study:

• To propose a novel and accurate camera calibration optimization method.
• To enhance the stability and robustness of camera calibration processes.

Main Methods:

• Combines convex lens imaging with the Wolf Pack Predation (WPP) bionic algorithm (CLI-WPP).
• Optimizes internal and radial distortion parameters using WPP.
• Utilizes reprojection error as the fitness criterion for WPP.
• Incorporates a reverse learning strategy to prevent local optima.

Main Results:

• The proposed CLI-WPP method achieved an average reprojection error of 0.06615037.
• Significantly lower error compared to simulated annealing (0.4298), Zhang's method (0.2884), Sparrow Search (0.2354), Particle Swarm (0.2193), and standard WPP (0.1063).
• Demonstrated superior calibration accuracy, stability, and robustness.

Conclusions:

• The CLI-WPP method offers a significant advancement in camera calibration accuracy.
• The integration of reverse learning effectively addresses the local optimum problem in WPP.
• CLI-WPP outperforms existing common optimization algorithms for camera calibration.