Scalable Training of Classification Systems Using Heterogeneous Input Data

Effective monitoring is an important component in safety and security related applications. Equipment failure, trespassing, theft and vandalism are all regular occurrences that the companies must deal with. These kinds of problems emphasize the importance of designing a state-of-the-art system for monitoring. In this project a system will be devised and trained to detect different moving-objects (pedestrians at the first step) and classify them into different groups. The system will be trained steadily by adding missing targets to the initial database. This intelligent classification will make monitoring systems to be capable of producing reports (It can be about number of people, vehicles in the site) and setting an alarms when it is necessary (For instance in case of animal detection). This work will allow company to develop a unique system for remote video monitoring.

Elmira Amirloo Abolfathi
Faculty Supervisor: 
Dr. Kyle O'Keefe