OPINION / Intelligent video: Making it work in outdoor environments
By John Romanowich
Video analytics has been shown to work well in controlled environments, such as those indoors. However, outdoor applications have raised the bar due to the considerable environmental variations and substantially larger areas of intended coverage.
Too often, outdoor intelligent video surveillance applications have created such a high rate of nuisance alarms and mis-detects that the system loses all security credibility. Technologies are now available that maintain a high probability of detect, while addressing the core issue of nuisance alarms by eliminating the environmental variations through image processing in advance of providing video analysis.
The reason analytics have been disappointing when applied outdoors is because the environment is much more volatile than indoors. Consider a typical outdoor scene: Cameras are mounted on high poles where even slight winds cause shaking, moving the entire scene. Clouds are moving across the sky, creating shadows on the ground that appear to the camera as moving objects. Trees and their leaves will flap in the breeze, further creating the appearance of moving objects. Water in the scene (which causes reflection from waves or puddles) only worsens the situation. When you add in snow, rain, humidity and dust, such a dynamic environment will wreak havoc for analytics not intended to operate under such conditions.
Making intelligent video effective for outdoor applications requires the use of technology and packaging specifically designed to address these outdoor variables. This begins with sufficient on-board camera processing power to provide image processing in advance of the analytics, which first stabilizes the image for camera motion, adapts to variables such as changing lighting, fog, rain, snow and sandstorms, and filters variables such as small animals, blowing debris, trees moving in the breeze and reflections from water.
Such processing power can also be engineered to communicate directly with the imager within the camera itself, making 100 percent of the raw uncompressed video available for analysis, greatly improving the probability that these cameras will accurately detect targets and filter the outdoor impediments that would otherwise trigger nuisance alarms. When video analysis is performed by video encoders separate from the camera -- or by servers in the datacenter -- they perform their analysis on a small fraction of the available scene information, at times less than one percent due to preparing data for transmission.
In the static well-lit indoors, where human targets are very large due to space constraints, such data loss has negligible impact on detection accuracy. However, in the outdoors, lighting, the background environment and target size vary substantially, and the loss of so much detail proportionally degrades the ability to accurately detect targets, or to properly distinguish nuisance alarms from legitimate targets.
The economics of covering large outdoor areas is also different. Outdoor surveillance involves additional infrastructure costs, including engineering design, construction, trenching, camera poles, network connectivity, video display and storage. By providing the appropriate level of computational power, such outdoor cameras are able to cover great distances, as much as three to five times the distance (more than 10 times the area) of indoor surveillance cameras, reducing the number of cameras, infrastructure and associated costs. While outdoor cameras may have a higher cost-per-unit, their extended range from extra processing leads to an overall reduction in deployment cost.
Effective perimeter security response and early detection systems for high-security applications -- such as government, transportation, energy, utilities or large campuses -- can be surprisingly cost-effective while at the same time, be trusted, secure and highly accurate. By applying the right technology to each application, such organizations can meet outdoor security objectives with much higher levels of accountability and cost effectiveness.
John Romanowich is CEO of SightLogix, Inc. He can be reached at: