A KD-tree and random sample consensus-based 3D reconstruction model for 2D sports stadium images

Area of Science:

• Computer Vision
• 3D Reconstruction
• Geometric Modeling

Background:

• 3D reconstruction of building images presents challenges in capturing fine details.
• Existing methods struggle with complex scenes like stadiums, leading to incomplete models.

Purpose of the Study:

• To develop an improved 3D reconstruction model for 2D building images.
• To enhance the accuracy and detail of reconstructed 3D building models, particularly for stadium environments.

Main Methods:

• Proposed a novel 3D reconstruction model integrating KD-tree and Random Sample Consensus (RANSAC).
• Utilized an improved KD-tree algorithm for enhanced matching rate in 2D image data extraction.
• Employed a screening method to convert sparse 3D models into denser representations.

Main Results:

• The integrated KD-tree and RANSAC algorithm demonstrated a superior matching rate for stadium scenes.
• The proposed method effectively increased the density of 3D models from sparse data.
• Simulation experiments showed significantly lower error rates compared to existing algorithms.

Conclusions:

• The proposed KD-tree and RANSAC-based 3D reconstruction model is highly suitable for building images.
• The method offers improved accuracy and detail for complex architectural scenes.
• This research advances the field of 3D reconstruction for architectural visualization and analysis.