Intruder Detection and Tracking Using Computer Vision and IoT

According to FBI Uniform Crime Reporting (UCR) statistics per year, there are roughly 2.5 million property-related burglaries, and out of which, the police solved only 13% of these reports due to the lack of a witness. Traditional methods such as using a CCTV surveillance system require a person to constantly monitor those cameras. This may be an effective way in locations such as offices and other public places. However, it cannot be used in houses as it would not only breach the privacy of the home owners but also not an effective way to detect an intrusion. In this paper, we propose eliminating the human aspect of monitoring cameras by implementing an autonomous intruder detection and tracking system. This system contains an indoor unit and an outdoor unit, and these two units communicate with each other using TCP/IP sockets. The indoor units contain a single camera present inside the house which uses face recognition to detect intruder. When the intruder leaves the house, the outdoor unit which contains multiple cameras located outside the house uses the image of the intruder taken by the indoor unit to look for the intruder. These two units use SMTP protocol to send email alerts when the intruder is detected. The indoor unit autonomously detects the intruder, while the outdoor unit autonomously locates the intruder. The salient feature is that by implementing this system, we can offset the lack of presence of a physical witness and aid the police in catching the intruder.