Future Predicting Intelligent Camera Security System

Abstract

The need for continuous human supervision, i.e. these systems are unable to perform certain functions without any operator monitoring the cctv video, is one of the major disadvantages of any camera-based monitoring system. An individual can hold his/her focus on the screen feed for a limited number of hours and not be distracted. Such poor monitoring could lead to a reduction in the effectiveness of the human operator's immediate action against a potential threat if detected on the screen. Therefore, the unique features of future prediction through the live video analysis method would have a huge effect on the surveillance system-based industries in order to address the limitations of the human attention span. The proposed system would be able to process the live stream from the cctv camera and generate output that will warn the operator of any possible danger that appears to occur or is occurring. For this method, a deep learning architectural approach is used with Convolutional Neural Networks. In this way, the surveillance camera system can detect an individual and identify and recognize those items carried by him or her on the basis of the level of danger. Similarly, the system can also identify and classify those acts that occur in the video feed into three distinct categories: natural, suspicious, malicious (based on the threat level) and send an alert to the respective human operator. Thus, the system will be able to help companies overcome security surveillance challenges and protect themselves from theft or any act of violence taking place in the area surrounding cctv.