Simplification method for 3D Terracotta Warrior fragments based on local structure and deep neural networks

Area of Science:

• Archaeology
• Computer Science
• Digital Heritage

Background:

• Three-dimensional (3D) scanning facilitates digital preservation and online presentation of cultural heritage artifacts.
• Terracotta Warrior fragments often present challenges due to redundant points in raw 3D scans, leading to large storage needs and processing times.
• Effective pre-processing of 3D fragment data is crucial for efficient digital archiving and analysis.

Purpose of the Study:

• To propose an effective method for simplifying 3D point clouds of Terracotta Warrior fragments.
• To reduce storage space and post-processing time for fragmented archaeological artifacts.
• To preserve essential geometric features during the simplification process.

Main Methods:

• An algorithm was developed to extract feature points from point clouds based on local structure using k-dimension trees.
• Feature and non-feature points were separated by comparing a feature discriminant parameter with a characteristic threshold.
• A deep neural network was employed to simplify the non-feature points, followed by merging with feature points.

Main Results:

• The proposed method successfully simplified 3D point clouds of Terracotta Warrior fragments.
• Experiments demonstrated excellent simplification results on both public and real-world datasets.
• The geometric features of the fragments were well-preserved after simplification.

Conclusions:

• The developed point cloud simplification technique is effective for 3D Terracotta Warrior fragments.
• This method significantly enhances the efficiency of digital storage and post-processing of archaeological data.
• The approach offers a valuable tool for the digital preservation and study of fragmented cultural heritage.