Reassembly of archeological artifact fragments requires considerable time and manual effort for researchers along with a necessity to handle potentially fragile material. It would therefore be advantageous for this process to be automatic. In this project we will work on creating a method for matching 3D scans of ancient animal bone fragments in order to reconstruct them. This problem can be viewed as a puzzle solving problem, for which there has been some interesting work in recent years [2]. This is a challenging task which will require the development/application of an unsupervised machine learning method to perform non-overlapping 3D point registration [3].
This is a challenging project for a motivated student with a keen interest in machine learning/optimisation and computer vision. There will also be an opportunity for the student to participate in a research paper as part of the wider project.
References:
[1] Papaioannou G, Schreck T, Andreadis A, Mavridis P, Gregor R, Sipiran I, Vardis K. From reassembly to object completion: A complete systems pipeline. Journal on Computing and Cultural Heritage (JOCCH). 2017 Mar 21;10(2):1-22.
[2] Paumard MM, Picard D, Tabia H. Deepzzle: Solving visual jigsaw puzzles with deep learning and shortest path optimization. IEEE Transactions on Image Processing. 2020 Jan 7;29:3569-81.
[3] Hu, Pengpeng, and Adrian Munteanu. “Method for registration of 3D shapes without overlap for known 3D priors.” Electronics Letters 57.9 (2021): 357-359.