Smart Home Surveillance System Through Edge Computing
Open Access
- Author:
- Wong, Joseph C
- Area of Honors:
- Computer Engineering
- Degree:
- Bachelor of Science
- Document Type:
- Thesis
- Thesis Supervisors:
- Sencun Zhu, Thesis Supervisor
Dr. Vijaykrishnan Narayanan, Thesis Honors Advisor - Keywords:
- smart home
machine learning
deep learning
motion detection
facial recognition
security
edge computing - Abstract:
- Household surveillance systems that aim to provide real-time alerts for security threats face the issue of lacking the computational power to detect meaningful events while delivering the alerts with low latency. Many home security cameras focus mainly on only capturing video for subsequent playback. If a security camera wants to be able to carry out more worthwhile tasks that are significant to the homeowner (e.g. alerts other than for motion detection) in a timely manner, it must be robust enough to do so. While mobile devices can be built to retain enough power for completing demanding jobs, such as prediction using deep learning, the development and production of these cameras may be very expensive and not feasible for the average household to use. This research utilizes the concept of edge computing to develop a low-cost household surveillance system design that increases security and privacy, largely focusing on a smart camera. This new system aims to be able to detect motion, identify known or unknown people, and lastly, detect delivery package theft and provide a prompt alert upon detection of an interesting event. The image processing and machine learning techniques used to achieve this system are not original techniques, however, the main contribution of this research derives from the design choices of this system of providing a low-cost mobile security camera that can ensure punctual alerts.