Effective Detection of Weapons in Video Surveillance

Abstract

Surveillance cameras also known as  Closed Circuit Tele-Vision (CCTV) play a major part in  the enforcement of the law and the administration of  justice. The present level of industrial production  technology cannot differentiate between a good and a  bad, or even a bit worse, situation. As a consequence,  additional crime scene investigation or law order  maintenance work is needed, which takes a lengthy time.  The suggested work is being utilized for a number of  purposes, including surveillance, weapon monitoring and  classification, live tracking, and more. Video input is  permitted as a type of raw input in this project for  monitoring and detecting abnormal events utilizing real time detection techniques such as You Look Only Once  Version 3 (YOLO V3). The proposed project's operations  make use of a processing module for object recognition  using convolutional neural networks such as YOLO V3,  which predicts classes and bounding boxes for the entire  image in a single run of the algorithm. The circular area  will be watched by CCTV, which will automatically  execute all operations and be controlled. Before  implementing in such a setting and delivering optimal  results, numerous samples and datasets will be examined  to find accuracy in detection and classification. The  planned effort aims to significantly reduce crime rates  while simultaneously providing improved protection in  specific areas and decreasing the time it takes to capture a  criminal.