There are many work done in this area even back to 20 years ago. But people are still working in this area. Most of them are computer vision researchers, working on the detection and tracking algorithms.
"Sensor Placement for Effective Coverage and Surveillance in Distributed Sensor Networks" aimed at optimizing the number of sensors and determining their placement to support distributed sensor networks. The optimization framework is probabilistic due to the uncertainty associated with sensor detections. The proposed algorithms are greedy. So good news is that we can try our placement algorithms in the setting of video surveillance, but the problem is the three cameras are already mounted on the roofs and it is not easy to move them or place a few more.
Then, what can we do with these three cameras? Try out detection and tracking algorithms?