Xavier Fernando
Software / Intelligent Instrumentation
According to Securehouse.Ca there are 159338 burglaries reported in Canada in 2015 alone. That means one out of every 28 houses are burglarized. This is unacceptable. Although many houses have installed surveillance cameras to prevent burglaries, the cameras are not effective when it is dark. Can we do advanced video processing to automatically identify thieves before a burglary and alert the officials. This is the objective of this project.
To develop an object identification and tracking algorithm in low light video environment that should alert the authorities.
Video processing in dark environments is in its preliminary stages. Nevertheless, it has enormous potential, since nowadays many buildings are equipped with video surveillance systems. Some work has been done in detecting and tracking human movements in videos. However, this is very difficult in low-light artificial illumination conditions with uneven lighting distribution. There are mainly black, or gray tones at night times and often the intruder’s clothes are dirty and dark colored which easily blends with the background in low illumination.
This project requires extensive computing. The students will be given an account in Southern Ontario Smart Computing Platform (SOSCIP). They can run deep learning algorithms in the powerful SOSCIP platform for object identification in low light environments. This is basically a Linux environment. I do have a Post Doctoral Fellow working in this area. He will guide the students to get started in supercomputing platform.
The group is responsible for the successful completion of the overall project.
Deep learning algorithms
Video Signal Processing
Object tracking in low light environment
XF02: BURGLAR DETECTION IN DARK | Xavier Fernando | Not yet submitted at No time