Solving Double-Sided Puzzles: Automated Assembly of Torn-up Banknotes Evidence

Area of Science:

• Computer Vision
• Forensic Science
• Digital Archiving

Background:

• 2D object reconstruction is crucial in fields like forensics, archiving, and banking.
• Existing literature often simplifies reconstruction as a one-sided puzzle problem.
• Torn banknote reconstruction presents unique challenges due to its double-sided nature.

Purpose of the Study:

• To address the problem of torn banknote reconstruction as a novel double-sided puzzle.
• To introduce a new dataset, the Torn Banknote (ToB) dataset, for this task.
• To develop and evaluate a reconstruction system utilizing keypoint-based methods.

Main Methods:

• A Borda count-based selection approach was employed to identify the optimal keypoint-based method.
• The Accelerated-KAZE (AKAZE) algorithm was selected as the most effective method.
• New metrics were proposed to quantify the success and loss ratios of reconstructed banknotes.

Main Results:

• The Accelerated-KAZE (AKAZE) method demonstrated superior performance in torn banknote reconstruction.
• The proposed reconstruction system achieved an average success rate of 95.55% based on the new metric.
• The study successfully created and utilized the Torn Banknote (ToB) dataset.

Conclusions:

• Torn banknote reconstruction can be effectively modeled as a double-sided puzzle problem.
• The developed AKAZE-based system offers a highly successful solution for banknote reconstruction.
• The proposed metrics provide a reliable way to evaluate the accuracy of reconstructed banknotes.