Violence Detection in Video Using Spatio-Temporal Features

Computer Analysis of Images and Patterns, 2011

Whereas the action recognition community has focused mostly on detecting simple actions like clapping, walking or jogging, the detection of fights or in general aggressive behaviors has been comparatively less studied. Such capability may be extremely useful in some video surveillance scenarios like in prisons, psychiatric or elderly centers or even in camera phones. After an analysis of previous approaches we test the well-known Bag-of-Words framework used for action recognition in the specific problem of fight detection, along with two of the best action descriptors currently available: STIP and MoSIFT. For the purpose of evaluation and to foster research on violence detection in video we introduce a new video database containing 1000 sequences divided in two groups: fights and non-fights. Experiments on this database and another one with fights from action movies show that fights can be detected with near 90% accuracy.

Advances in Data and Information Sciences

There are different methods or techniques used for identifying violence from video, such as hitting some object, kicking, fighting, and punching someone but still there is a big challenge for us to identify violence. However, some of the earlier mechanism generally extract descriptors around the spatiotemporal interesting points (STIP) or extract statistic features but there is limited effectiveness in detecting videobased violence. Therefore, the objective is to develop a better violence identification system that identifies the violence and triggers an alarm so that prompt assistance will be provided. This paper helps researchers who wish to study violent activity recognition and gather different insights on the main challenges and issues to solve in this emerging field.

International Journal for Research in Applied Science and Engineering Technology IJRASET, 2020

The aim of the project is to come up with alarm just in case of abnormal activities and to help human operators and for offline review. A challenge is to detect/develop intelligent video systems capable of mechanically analyzing and investigate violence within the scene. Such capability could also be extraordinarily helpful in some videos police investigation eventualities like in prisons, medical specialty centers or perhaps in sensitive areas within the town. I. INTRODUCTION In the last years, the matter of act recognition at a distance has become tractable by CV techniques. Though the primary approaches obtained sensible results, they have some limitations too. There are, as an example, aperture issues and discontinuities in optical flow based mostly approaches, and illumination and re formatting issues in feature chase approaches. The goal of this paper is to assess the performance of contemporary action recognition approaches for the popularity of fights in videos, movies or video-surveillance footage. Most of previous work on action recognition focuses on straightforward human actions like walking, jumping or hand waving. Despite its potential quality, violent action detection has been less studied. Whereas there's variety of well-studied datasets for action recognition, vital datasets with violent actions haven't been created accessible. In the work we have introduced a fight dataset to assess the performance within the fight detection. A violence detector has immediate pertinence each within the police investigation domain and for rating/tagging online video content. The first perform of large scale police investigation systems deployed in establishments like faculties, prisons and elder care facilities is for alerting authorities to probably dangerous things. However, human operators are flooded with the quantity of camera feeds and manual response time is slow, leading to a robust demand for machine-driven alert systems. Similarly, there's increasing demand for machine-driven rating and tagging systems that may method the nice quantities of video uploaded to websites. The major contribution of this paper are twofold. First, we have shown that one will construct a flexible and correct fight detector employing a native descriptors approach. Second, we have used a new dataset of hockey videos containing fights and demonstrate that our projected approach faithfully notices violence in sports footage, even within the presence of camera motion.

Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, 2019

The automatic classification of violent actions performed by two or more persons is an important task for both societal and scientific purposes. In this paper, we propose a machine learning approach, based a Support Vector Machine (SVM), to detect if a human action, captured on a video, is or not violent. Using a pose estimation algorithm, we focus mostly on feature engineering, to generate the SVM inputs. In particular, we hand-engineered a set of input features based on keypoints (angles, velocity and contact detection) and used them, under distinct combinations, to study their effect on violent behavior recognition from video. Overall, an excellent classification was achieved by the best performing SVM model, which used keypoints, angles and contact features computed over a 60 frame image input range.

IEEE Access

With the rapid growth of surveillance cameras to monitor the human activity demands such system which recognize the violence and suspicious events automatically. Abnormal and violence action detection has become an active research area of computer vision and image processing to attract new researchers. The relevant literature presents different techniques for detection of such activities from the video proposed in the recent years. This research study reviews various state-of-the-art techniques of violence detection. In this paper, the methods of detection are divided into three categories that is based on classification techniques used: violence detection using traditional machine learning, using Support Vector Machine (SVM) and using Deep Learning. The feature extraction techniques and object detection techniques of each single method are also presented. Moreover, datasets and video features that used in the techniques, which play a vital role in recognition process are also discussed. For better understanding, the steps of the research approaches have been presented in an architecture diagram. The overall research findings have been discussed which may be helpful for finding the potential future work in this research domain.

International Journal of Latest Research in Engineering and Technology (IJLRET), 2016

The demand for automatic action recognition systems have increased due to the rapid increase in the number of video surveillance cameras installed in cities and towns. Automatic action recognition system can be effectively used to generate on-line alarm in case of abnormal activities to assist human operators and for offline inspection. Although action recognition problem has become a hot topic within computer vision, detection of violent scenes receives considerable attention in surveillance system which is justified by the need of providing people with safer public spaces. This survey discusses the current state of the art methods and techniques that are being applied for the task of automated violence detection.This survey emphasizes on motivation and challenges of this very recent research area by presenting approaches for violence recognition in surveillance video. This paper aims at being a driving force for researchers who wish to approach the study of violent activity recognition and gather insights on the main challenges to solve in this emerging field.

Proceedings of the 9th International Conference on Computer Vision Theory and Applications, 2014

Whereas the action recognition problem has become a hot topic within computer vision, the detection of fights or in general aggressive behavior has been comparatively less studied. Such capability may be extremely useful in some video surveillance scenarios like in prisons, psychiatric centers or even embedded in camera phones. Recent work has considered the well-known Bag-of-Words framework often used in generic action recognition for the specific problem of fight detection. Under this framework, spatio-temporal features are extracted from the video sequences and used for classification. Despite encouraging results in which near 90% accuracy rates were achieved for this specific task, the computational cost of extracting such features is prohibitive for practical applications, particularly in surveillance and media rating systems. The task of violence detection may have, however, specific features that can be leveraged. Inspired by psychology results that suggest that kinematic features alone are discriminant for specific actions, this work proposes a novel method which uses extreme acceleration patterns as the main feature. These extreme accelerations are efficiently estimated by applying the Radon transform to the power spectrum of consecutive frames. Experiments show that accuracy improvements of up to 12% are achieved with respect to state-of-the-art generic action recognition methods. Most importantly, the proposed method is at least 15 times faster.

Information, 2020

This benchmarking study aims to examine and discuss the current state-of-the-art techniques for in-video violence detection, and also provide benchmarking results as a reference for the future accuracy baseline of violence detection systems. In this paper, the authors review 11 techniques for in-video violence detection. They re-implement five carefully chosen state-of-the-art techniques over three different and publicly available violence datasets, using several classifiers, all in the same conditions. The main contribution of this work is to compare feature-based violence detection techniques and modern deep-learning techniques, such as Inception V3.

Data in Brief, 2020

Challenges and Methods of Violence Detection in Surveillance Video: A Survey, 2019

This article presents a survey of the latest methods of violence detection in video sequences. Although many studies have described the approaches taken to detect violence, there are few surveys providing exhaustive review of the available methods. We expose the main challenges in this area and we classify the methods into five broad categories. We discuss each category and present the main techniques that proposed improvements as well as some performance measures using public datasets to evaluate the different existing techniques of violence detection.