An Efficient Criminal Segregation Technique Using Computer Vision

Abstract

In the contemporary world, where the population has been growing rapidly, it has become difficult to identify suspicious persons. Given the abundance of population in public places, it is difficult to identify a culprit post-crime activity because one (in general, the investigator) has to go through the entire CCTV footage to track and pin down people who seem suspicious for further investigation. These traditional methods are very time-consuming and laborious since each footage can be at least hours long. This proposed method takes advantage of the fact that the culprit tries to hide their identity by either evading the camera or by masking themselves. In places like shopping malls, movie theaters, restaurants, etc. these cameras are placed at the entrance and at security checks. Hence, it is not plausible for them to completely evade the cameras. This shifts our concentration to the latter idea that they hide their identity by masking themselves. We build our model on this flaw and combine video surveillance with machine intelligence to provide an efficient interface than unprocessed video feed. Furthermore, this system is not only useful for post-crime scenarios but can also be deployed for real-time analysis.