Data-Weighted Multivariate Generalized Gaussian Mixture Model: Application to Point Cloud Robust Registration

Area of Science:

• Computer Vision
• Computational Geometry
• Machine Learning

Background:

• Point cloud registration is crucial for 3D scene reconstruction and analysis.
• Existing methods often struggle with noise, outliers, and varying data density.
• Robust and accurate registration algorithms are needed for real-world applications.

Purpose of the Study:

• To propose a novel weighted multivariate generalized Gaussian mixture model for point cloud registration.
• To enhance registration accuracy and robustness against noise and outliers.
• To develop an efficient stochastic optimization approach for parameter estimation.

Main Methods:

• Utilizing a weighted multivariate generalized Gaussian mixture model.
• Employing the Expectation-Maximization (EM) algorithm with a fixed-point method for parameter updates.
• Determining the number of components using the Minimum Message Length (MML) criterion.
• Applying KL divergence as a loss function for stochastic optimization.
• Evaluating performance on self-built point clouds for rigid registration.

Main Results:

• The proposed algorithm significantly reduces the impact of noise and outliers.
• Effective extraction of key features from data-intensive regions is achieved.
• Demonstrated robust performance in rigid point cloud registration.
• The method shows promise for accurate 3D scene matching.

Conclusions:

• The weighted multivariate generalized Gaussian mixture model combined with stochastic optimization offers a robust solution for point cloud registration.
• The algorithm's ability to handle noisy data and extract salient features makes it suitable for complex 3D environments.
• This approach advances the state-of-the-art in 3D point cloud processing and registration.