Investigation of Bite Marking Detection and Classification using Segmentation with Deep Transfer Learning Approach

Abstract

Bite marking is considered an important topic in forensic investigation. The detection and classification of bite marking in the victim will be helpful to properly determine the criminals. The recent advances in computer vision (CV) and artificial intelligence (AI) models have resulted in the design of bite marking detection and classification models. At the same time, the available machine learning (ML) and deep learning (DL) models pave a way for effective bite marking detection and classification outcomes. This study investigates a new bite marking detection and classification using segmentation with deep transfer learning (BMDC-SDTL) approach. The major intention of the BMDC-SDTL technique is to determine the appropriate class labels for the bite marked images. The BMDC-SDTL technique primarily involves different stages of pre-processing. In addition, the BMDC-SDTL technique designs a new Chan-Vese segmentation approach to identify the bite marked regions. Followed by, DenseNet-169 model is employed for feature extraction process. Finally, support vector machine (SVM) and logistic regression (LOR) models are utilized as classification models. The performance validation of the BMDC-SDTL technique is performed using a dataset collected by our own. An extensive comparison study reported the better outcomes of the BMDC-SDTL technique over the other techniques.