International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 09 Issue: 07 | July 2022
p-ISSN: 2395-0072
www.irjet.net
Weapon Detection and Classification in CCTV Footage Tejashwini PS1, Saatvik Girish Nargund2, Sumuk K2, Rohith G R2, Trinetra BM2 1Assistant
Professor, Department of CS&E, BIT, Bangalore, Karnataka, India Department of CS&E, BIT, Bangalore, Karnataka, India ---------------------------------------------------------------------***--------------------------------------------------------------------2Student,
Abstract - Closed Circuit Television (CCTV) cameras are
weapons which may be present in the CCTV footage. Weapons can include knife and pistols. Whenever a weapon is detected, it will alert the user (CCTV operator). It will also be possible to add exceptions to the rules if the user wishes.
being used for surveillance and to monitor activities i.e. robberies, but these cameras still require human supervision and intervention. We need a system that can automatically detect these illegal activities. This work focuses on providing a secure place using CCTV footage as a source to detect harmful weapons by applying the state of the art open-source deep learning algorithms. We have implemented a classification model with different classes of weapons and relevant confusion objects inclusion concept is introduced to reduce false positives and false negatives
2. RESEARCH ELABORATION We have researched various papers to get to understand the different methods used by various authors in their own weapon detection systems. Aarchi Jain et. al. explored the usage of gun detection using the Haar Cascade Classifier [1]. The low accuracy on pistols was a drawback in this system. Harsh Jain et. al. implemented a system using a CNN based SSD algorithm, which offered improved precision [2]. Pawel Donath et. al. presented a neural network based system using a custom neural net inspired by AlexNet and VGGNet architectures [3]. Michal Grega et. al. have used descriptors like edge histogram and homogenous texture, along with SVM, in their system to detect pistols and knvies [4]. JLS Gonzalez, in their proposed system, have used faster RCNN, along with creation of synthetic datasets using Unity game engine to compensate for the lack of good quality date.
Key Words: Machine Learning, Convolutional Neural Network, Weapon Detection, Pistol Detection, Knife Detection
1. INTRODUCTION Security and safety is a big concern for today’s modern world. For a country to be economically strong, it must ensure a safe and secure environment for investors and tourists. Having said that, Closed Circuit Television (CCTV) cameras are being used for surveillance and to monitor activities i.e. robberies. It is considered as one of the most important evidence in law enforcement agencies and courts. Therefore, the amount of CCTV cameras installed have increased in number all over the world. This has made the public feel much safer, and decreased crime in many areas all over the world. As a result of the increase in number of CCTV cameras, the number of screens that have to be monitored by a single CCTV operator has increased a lot. One operator cannot be expected to monitor many screens at a single time. Also, he cannot constantly monitor the footage throughout the day Moreover, it is difficult and expensive to monitor hundreds, or even thousands of CCTV video footage in an area. Therefore, there is an increasing demand to automate CCTV surveillance.
Muhammad Tahir Bhatti et. al. implemented and tested various models for Pistol Detection in CCTV footage [6]. All the models tested were based on convolutional neural networks (CNN). Kushagra Yadav et. al. used N stage learning, along with ADADELTA technique to train a neural network to detect different types of weapons and knives [7]. Jacob Rose et. al. propsed a Faster RCNN based model using the ResNet – 50 base network for weapon detection [8]. This model was trained on COCO dataset. Mitchell Singleton utilized the MobileNetV1 Neural Network to identify handguns in in various orientations, shapes, and sizes [9]. Jesus RuizSantaquiteria et. al. combined, in a single architecture, both weapon appearance and human pose information to better detect weapons in an image [10].
3. METHODOLOGY
1.1 Objectives of the system
The techniques used in this paper mainly center around object recognition. It is a complex task which actually involves 3 sub tasks: Image Classification, Object Localization, and Object Detection. Image Classification predicts the class of an object in an image. Object Localization locates the presence of objects in an image and indicate their location with a bounding box. Object Detection locates the presence of objects with a bounding box and detect the classes of the located objects in these boxes.
The main aim of automated CCTV surveillance is to alert the CCTV operator whenever there is a dangerous situation. A dangerous situation refers to a person or a group of people attacking, creating fear or disturbances with weapons like knives and guns. These kinds of situations can be detected by automated systems. The systems work by using object detection algorithms to classify objects detected in the video footage. Our proposed system will detect and classify any
© 2022, IRJET
|
Impact Factor value: 7.529
|
ISO 9001:2008 Certified Journal
|
Page 1869