Video Surveillance at Night

Stevens, M. R., Pollak, J., Ralph, S., and Snorrason, M.

Proceedings of WACV, Breckenridge, CO (January, 2005)

In this paper, we focus on the problem of automated parking lot surveillance at night using visible-and-nearinfrared cameras (VNIR). These sensors are low cost (hundreds of US dollars) when compared with thermal infrared sensors (thousands of dollars) and have a much higher resolution. Thermal imaging may seem an obvious choice for nighttime surveillance since it enables detection of people and warm cars without any ambient light. However, most parking lots have illumination levels between 5 and 50 lux even on moon-less nights. In fact, parking lots next to buildings with strict security tend to be well lit as a deterrent to illegal activity. Intuitively, cameras optimized for low-light are therefore just as appropriate as thermal cameras for this scenario. Long-wave infrared (LWIR) 160×120 cameras are the least expensive thermal solution, but they still cost anywhere from $7,000 to $15,000. Finally, the ability of thermal cameras to detect cars generally depends on the engine being warm, hence a cold car starting up and driving away can be difficult to image. We call our research system VANESSA, for Video Analysis for Nighttime Surveillance and Situational Awareness. VANESSA is capable of: 1) enhancing a video stream, 2) determining moving objects, and 3) rejecting false motion due to vehicle headlights. This paper focuses on identifying the key difficulties associated with our chosen domain and camera technology. In addition, we present algorithms to correctly detect and segment moving objects while maintaining a low false alarm rate. This is a non-trivial task given the difficult nature of the imagery.

For More Information

To learn more or request a copy of a paper (if available), contact

(Please include your name, address, organization, and the paper reference. Requests without this information will not be honored.)