Developing Intelligent Video Algorithms for Surveillance Applications
Hi. I am Peter Tu. I am a computer scientist here at GE Global Research, working on intelligent video programs to drive technology development for GE Security’s video surveillance product applications.
We are developing a comprehensive systems approach to video surveillance. The goal is to provide real-time site awareness over large networks of surveillance cameras. A state-of-the-art video test-bed has been constructed on site at our research facility in Niskayuna with 8 permanently installed pant tilt zoom cameras and a computing cluster that enables real time processing and meta-data visualization.
Our lab is pursuing a number of research themes, which I will introduce on my blog as we go forward. Today, I wanted to start things off by showing you some cool video from our video test-bed, which demonstrates our ability to track multiple people across multiple camera views and present the results in a synthetic One World View. Just click on the picture at left to see the video.
The goal is to provide a security system that is always vigilant and always watching. The more we can introduce automated features like this into the system that can see more and spot suspicious behavior, the more vigilant these systems can be.
1. The picture of the two persons shown is not clear. Can the camera distinguish between them or recognize them more clearly so as to dientify who the persons are?
2. What is the maximum distance that these cameras can view the images clearly?