Symbiotic Organism Search Optimization with Deep Learning based Bite Mark Identification in Forensic Dentistry

• Published in: 2023 6th International Conference on Engineering Technology and its Applications (IICETA) pp. 856 - 862
• Main Authors: Alkhafaij, Mahdi abdulkhudur, Oleiwi, A. Sahib, Ali, Rabei Raad, Ali, Eyhab, Almusawi, Muntather, Al-Tahee, Mustafa
• Format: Conference Proceeding
• Language: English
• Published: IEEE 15-07-2023
• Subjects: Bite marking; Deep learning; Dentistry; Feature extraction; Forensic analysis; Forensic dentistry; Forensics; Metaheuristics; Organisms; Symbiosis; Teeth
• ISSN: 2831-753X
• DOI: 10.1109/IICETA57613.2023.10351436

Generally, forensic dentistry overcomes the challenge of recognizing individuals based on some particular features of bitemark or teeth impression. Bite mark identification usually has human bias and includes human interaction. It could be useful to have a system that has higher accuracy matching performance and decreases human bias. Many image analysis techniques like Image pre-processing, Image Acquisition, Feature extraction, Segmentation, and so on are used for identifying humans in lower time and also saves the investigation time. In criminal court, Bitemarks is considered the effective form of dental evidence in rape cases. Therefore, the study uses a new Symbiotic Organism Search Optimization with Deep Learning based Bite Mark Identification (SOSODL-BMI) technique in Forensic Dentistry. In the presented SOSODL-BMI technique, the bite mark of a person can be accurately identified in forensic investigations. Firstly, the Weiner filtering (WF) technique is used to preprocess the input bite mark images. Then, the residual network (ResNet) model is applied by the SOSODL-BMI technique to generate feature vectors with Adam optimizer. Finally, the artificial neural network (ANN) model is exploited for bite marking identification, and the SOSO algorithm performs the parameter tuning process. A detailed simulation analysis takes place on the bite marking image dataset to exhibit the enhanced performance of the SOSODL-BMI approach. The complete results displayed the improvements of the SOSODL-BMI approach over other recent methods.