BTech 451
Motion Detection for Video Surveillance
Commercial off-the-shelf video surveillance cameras are very capable of recording images in computer hard disks at high resolutions (such as 1920 x 1024) and frame rates (such as 30 frames per second). They are also capable of applying motion detection criteria to achieve surveillance objectives while reducing the bandwidth burden for data transmission and recording in computer storage. However the detection criteria are based on change of pixel counts and not on the meaning of images or the environment seen or recorded. This is not accurate and leads to lots of false alarms. Computer data modelling comes in at this point. Data modelling will attempt to give meanings to pixels of images captured by the camera, and allow meaningful actions to be taken.
We expect the above described data model to require a lot of computation power to work out the solutions. More to the point, the computation has to be done in real time so that detection is revealed in real time. We will test the algorithms with serial CPU cores and parallel GPU cores.
We expect the above described data model to require a lot of computation power to work out the solutions. More to the point, the computation has to be done in real time so that detection is revealed in real time. We will test the algorithms with serial CPU cores and parallel GPU cores.