Received 04.04.2022, Revised 27.06.2022, Accepted 25.07.2022

Prospects of neural network approach to the problem of damaged papers restoration

Maksym Solonyi, Andriy Yarovyi, Yaroslav Ivanchuk, Volodymyr Ozeranskyi

Nowadays, the damaged papers restoration task is quite urgent, experts spend hours, days or even weeks to restore damaged documents, drawings or other materials that can play the key role evidence in a criminal case. Automation of this process will significantly increase the speed and quality of solving this problem, thereby increasing the efficiency of the work of forensic experts. During the search for existing solutions, no direct analogs were found, but several indirect analogs were found that solve quite similar problems. The first analog is a technology proposed by Haifa University scientists to restore damaged archaeological finds. This technology was successfully tested on real artifacts of the British Museum, which proved its effectiveness in restoring damaged frescoes. These results are promising for the further development of information technology for restoring the integrity of a damaged document, in particular, in the context of the complete restoration of the paper structure based on its microrelief. The second analog is image editing technology using Kohonen maps. This technology effectively performs the basic tasks of retouching images, in particular, removing objects, restoring integrity after removal. Since this technology is used for image processing, it can be used as a basis for restoration of the damaged content of a document after its physical assembly. After all, during the paper structure restoration, the integrity of the content may be partially lost. In this article, each of the above technologies is analyzed in detail, including at the level of mathematical models, their advantages and disadvantages are highlighted, and examples of their real application are given. Based on the advantages of each of the analyzed technologies, an approach to solving the problem of damaged papers restoration is proposed.

neural networks, damaged papers restoration, criminalistics
55-60
Solonyi, M., Yarovyi, A., Ivanchuk, Y., & Ozeranskyi, V. (2022). Prospects of neural network approach to the problem of damaged papers restoration. Information Technologies and Computer Engineering, 19(2), 55-60. https://doi.org/10.31649/1999-9941-2022-54-2-55-60

References

[1] Soloniy, M.A., & Yarovy, A.A. (2021). Prospects of using neural network technologies for the problem of reproducing damaged papers. In Youth in science: research, problems, prospects (MN-2021): Proceedings of the All-Ukrainian scientific and practical conference (pp. 1-2). Vinnitsa: VNTU.

[2] Soloniy, M.A., & Yarovy, A.A. (2022). Analysis of intellectual technologies in the context of the problem of reproduction of damaged papers. In Youth in science: research, problems, prospects (MN-2022): Proceedings of the All-Ukrainian scientific and practical conference (pp. 1-3). Vinnitsa: VNTU.

[3] Derech, N., Tal, A., & Shimshoni I. (2018). Solving archaeological puzzles. Retrieved from https://arxiv.org/pdf/1812.10553.pdf.

[4] Eklavya. (2019) Kohonen self-organizing maps: A special type of Artificial Neural Network. Retrieved from https://towardsdatascience.com/kohonen-self-organizing-maps-a29040d688da.

[5] Self organizing maps – Kohonen maps. (2022). Retrieved from https://www.geeksforgeeks.org/self-organising-maps-kohonen-maps.

[6] De Castro, A.P.A., Drummond, I.N., & da Silva J.D.S. (2008). A multiscale neural network method for image restoration. Trends in Computational and Applied Mathematics, 9(1), 41-50. doi: 10.5540/tema.2008.09.01.0041.

[7] Favorskaya, M., Jain, L.C., & Bolgov, A. (2014). Image inpainting based on self-organizing maps by using multi-agent implementation. Procedia Computer Science, 35, 861-870. doi: 10.1016/j.procs.2014.08.253.