Execution of coordinate based classifier system to predict specific criminal behavior using regional multi person pose estimator

Abstract

There are numerous numbers of issues in society, one of which is crime. While crime refers to a wide range of deliberate, unlawful behaviors, the most archetypal ones involve murder, threatening and violent activities. Its expenditures and consequences affect almost everything but to a certain extent. If we want to prevent crime, we must first identify criminal activity. It is hard to locate unlawful behavior without a lot of effort. With crime surging at an alarming rate, several methods have been developed in the past to predict and prevent criminal activities. However, the methods available currently are not efficient enough to predict the extensive variety of criminal activities that occurs in modern days. To do so, we need to make greater use of technological advancements in order to forecast crime. This paper presents an approach that will be able to detect and predict crime by combining machine learning with a coordinate-based approach. The proposed apparatus integrates existing video footage to detect and analyze human behavior. The system distinguishes between human stances present in the scene in order to detect criminal behavior and subsequently predict crime. Using video processing, the methodology compares human stances with a trained dataset and detects those body positions that may indicate criminal activity.