Digital Reconstruction of Fragmented Artifacts Using Computer Vision Techniques

EasyChair Preprint 14255

Abstract

The preservation and analysis of fragmented artifacts present significant challenges in archaeology, art history, and cultural heritage conservation. Traditional methods for reconstructing these artifacts often involve labor-intensive manual efforts, which can be prone to error and lack scalability. Recent advancements in computer vision and digital imaging offer innovative solutions for this problem. This paper explores the application of computer vision techniques to the digital reconstruction of fragmented artifacts, focusing on methods such as 3D modeling, image registration, and machine learning algorithms.

We present a comprehensive review of existing approaches, including structure-from-motion (SfM), multi-view stereo (MVS), and deep learning-based segmentation. Each technique's strengths and limitations are discussed, and case studies are provided to illustrate their practical applications. Additionally, we introduce novel methodologies that leverage generative adversarial networks (GANs) and convolutional neural networks (CNNs) to enhance the accuracy and efficiency of reconstruction processes.

Our findings demonstrate that integrating computer vision with traditional archaeological methods not only accelerates the reconstruction process but also provides more precise and detailed models of fragmented artifacts. This integration facilitates better visualization, interpretation, and preservation of cultural heritage objects, ultimately contributing to more effective research and education in the field.

Keywords: Digital reconstruction, fragmented artifacts, computer vision, 3D modeling, deep learning, machine learning.

Keyphrases: Technology, computing, digital reconstruction

@booklet{EasyChair:14255,
  author    = {Favour Olaoye and Chris Bell and Peter Broklyn},
  title     = {Digital Reconstruction of Fragmented Artifacts Using Computer Vision Techniques},
  howpublished = {EasyChair Preprint 14255},
  year      = {EasyChair, 2024}}