Violence Activity Detection Classification - A Review

Abstract

With the emerging trends of different automated surveillance systems for the security of people, activities like violence activity detection have become an active area of research. We observe several criminal and abnormal violent activities in our daily lives that need can be detected on spot to avoid a bigger violent event. To prevent from violence and different kind of harmful activity its need to work on accuracy of that automated surveillance. This work aims to provide a systematic literature review on state-of-the-art violent activity detection methods, datasets needed to develop those frameworks for training and testing and, identify gaps through discussion that can be filled with further proposed solutions. In this study, we have compared recent papers from 2017 to 2023 in this domain for their data classes, open-source availability and statistics. Further, the methodologies are divided into domains i.e. Dataset that used by previous researchers and get results to detect violence activity in public and private place with different accuracy, Real-time violent activity detection in video using traditional approaches, machine learning approaches and deep learning approaches. In the end, we have identified the issues and gaps in the existing literature that can create a potential difference for future researchers in this domain.