A Deep Residual Network-Based Approach to Reidentify Persons via Face Features Extraction from Video Frames

The rise of unwanted situations in today's environment has made video security more important. Video surveillance helps with anomaly identification and improves security in a number of areas, such as businesses, educational institutions, and city monitoring systems. The task of recognizing and re-identifying moving subjects, particularly those captured by many cameras, are extremely difficult and are made more problematic by things like low-quality video, inconsistent illumination, and congested spaces. A model database, a feature descriptor, and a classifier are three essential elements of digital image processing that are used in the process of identifying people in video. This document aims to give a general summary of the procedures utilized in video footage-based person re-identification (Re-ID) The arena of person Re-ID study includes a range of approaches and techniques. These methods include the identification of faces through feature recognition, the creation of detailed profiles through multiple component matching representations, and the use of multiple component dissimilarity representations to capture the subtleties of a person's appearance. Using deep residual networks is one noteworthy method that has shown to be remarkably effective in extracting face characteristics. After these traits are recovered, they may be used to compare and match people's faces between various video frames and camera angles. In addition, an innovative deep learning-based technique for person Re-ID using face traits is presented in the research. Furthermore, the research examines the main obstacles related to human Re-ID in video surveillance and considers possible ways to overcome these obstacles.