An Iterative Closest Points Algorithm for Registration of 3D Laser Scanner Point Clouds with Geometric Features

Area of Science:

• Computer Vision
• 3D Data Processing
• Geometric Algorithms

Background:

• The Iterative Closest Points (ICP) algorithm is standard for 3D point cloud registration.
• ICP requires accurate initial values and approximate registration to avoid local minima.
• These prerequisites are often unmet in real-world point cloud matching scenarios.

Purpose of the Study:

• To develop an improved ICP algorithm that overcomes limitations of initial value dependency.
• To enhance the robustness and convergence properties of 3D point cloud registration.
• To enable accurate point cloud matching without requiring precise initial alignment.

Main Methods:

• Proposed a novel Geometric Feature-based Iterative Closest Points (GF-ICP) algorithm.
• Utilized point cloud geometric features (curvature, surface normal, density) for correspondence search.
• Integrated geometric features into the ICP error function for registration.

Main Results:

• GF-ICP demonstrated improved convergence speed compared to standard ICP.
• The algorithm showed an expanded interval of convergence.
• Accurate registration was achieved without the need for a proper initial value.

Conclusions:

• The proposed GF-ICP algorithm effectively addresses the initial value problem in point cloud registration.
• Incorporating geometric features enhances the performance and reliability of ICP.
• GF-ICP offers a more robust solution for accurate 3D point cloud matching.