IMEG Corp

As early analog-based CCTV systems have evolved into networked IP video surveillance systems, the capabilities, feature sets, and intelligence of systems available on the market have increased exponentially. As analog cameras were replaced by IP cameras, transmission infrastructure moved from dedicated coax networks to standard structured cable data networks, and recording systems advanced from video tape recorders to digital video recorders (DVRs) and later network video recorders (NVRs). Throughout this evolution, video surveillance data has become more valuable to organizations’ security efforts.

Over recent years, the quantity of cameras deployed at a typical commercial or government facility has increased to meet the demand for greater and greater visibility– but the ability for humans to effectively monitor these systems has been overwhelmed by the huge amounts of video data captured. Security monitoring personnel cannot view the large number of video feeds and be expected to successfully pick out security incidents or suspicious behavior on a real-time basis or even to search for security events from thousands of hours of recorded video. This limitation spurred the development of software applications that can process large amounts of video data and extract valuable video information, freeing security staff to apply human intelligence to evaluate and assess incident video to identify actual security threats or activities. The use of computer processing power and software algorithms to automate the processing of large collections of video data is referred to as“video content analysis” or more commonly as “video analytics.”

Advanced intelligence

Video analytic products have evolved from early offerings using rules-based analytics to more intelligent systems made possible by advancements in artificial intelligence (AI) and deep learning algorithms. These improvements provide more complex and powerful video processing to analyze video files, identify security incidents, and tag these incidents for further review, resulting in more accurate and reliable information to the user.

Early analytic systems offered a basic set of event triggers such as video motion detection, video loss detection, virtual tripwire sensing, areas of interest, as well as “smart search” tools to review recorded video. Current systems offer much more complex and valuable analysis tools capable of searching based on physical description, behavioral characteristics, vehicle description, etc. Although there are numerous analytic search criteria available, some that are more valuable to security include:

Increased processing power at the camera

In addition to more intelligent software algorithms, significant increases in processing power on-board IP cameras has opened new opportunities to deploy analytic applications at the edge. In the past, camera electronics were discrete and designed exclusively to capture video signals and enhance the image quality transmitted from the camera. As surplus processing capacity has become available on newer IP cameras, some of the video analysis workload traditionally done on centralized analytics servers can be performed at individual cameras in the field.

"It is exciting to see the continued maturity of security video analytics as advancements in software development and processing power continue to evolve"

By pushing portions of the video analysis tasks out to the cameras, improvements can be achieved in network bandwidth utilization, reductions in video storage requirements, and reduced deployment costs. By allowing security events to be identified and flagged at the edge, more information related to the video image can be captured and stored in the form of enhanced metadata attached to the video data stream. 

Along with technological advances, exciting progress is being made by organizations such as the Open Security & Safety Alliance (OSSA) to develop a standardized framework to encourage the development of new applications that can be loaded on the camera.