printer friendly version

The Battle: Human vs Machine

The human brain has limited attention span capabilities. A 1999 study (Green, 1999) found that after 20 minutes, guards watching a video scene will miss up to 95% of all activity. Leveraging advancements in video pattern detection and video analytics technology addresses this issue, and has comprehensively evolved from being a strictly forensic tool into a powerful proactive solution.

Paired with high-definition imaging, HD analytics provides security operators with highly accurate alerts and clear image detail, enhancing their ability to effectively intervene and take action when an incident occurs.

The economics of human attention

Artificial intelligence pioneer, Herb Simon stated, “A wealth of information creates a poverty of attention.” Simon noted that most technology systems were focused on providing as much information as possible without taking the human attention span into consideration. Consequently, these systems provided a surplus of information to people, when what was needed were systems that filtered out irrelevant information, and highlighted items of interest (Simon, 1996).

What Simon was describing was the theory of attention economics; an approach to the management of information that treats human attention as a scarce commodity and a limiting factor in the absorption of information. The attention economics theory supports the creation of systems that take attention capabilities into consideration in their design, creating filters to ensure the first content a user is presented with is relevant and of interest.

The security risk of the human attention span

Based on the theory of attention economics, most security control centres and corresponding video surveillance systems today present security personnel with a wealth of information, leading to a poverty of attention. The aforementioned study (Green, 1999) showed a disturbing trend in operator performance:

1. Security operator performance degrades considerably after 20 minutes.

2. Security operators cannot effectively monitor multiple surveillance cameras and sensors.

3. Poor image quality accelerates this rate of degradation.

4. Viewing twice the number of cameras accelerates degradation by a factor of two.

The concept of video analytics technology is to present only the information that will require an operator’s immediate attention. However, the vast majority of these systems create a disproportionate amount of irrelevant information, contributing to operator confusion and inaction.

The evolution of video analytics

Video analytics has evolved across a series of three technologies:

1. Video Motion Detection (VMD) – any change from one frame to another is important.

2. Advanced Video Motion Detection (AVMD) – any change that deviates from a background model is important.

3. Advanced Video Pattern Analytics – any change that has a pattern of a known object type is important.

VDM is now a standard feature included in most new surveillance cameras, recorders and video management software packages. The VMD feature focuses on detecting any pixel movement from scene to scene based on a simplistic user-defined threshold. VMD is most effective in sterile and static environments, however the technology is limited in dynamic environments, resulting in high false alarm rates. Unfortunately, this high rate of false alarms leads directly to a rapid decrease in operator attention.

In response to this limitation, the industry progressed from VMD to AVDM. AVMD is based on background modelling, alerting on any change that deviates from an established background model. This technology focuses on monitoring a scene and using the data captured via complex manual calibration to identify moving objects. AVMD is effective when set up and calibrated correctly, yet is limited when background composition changes (e.g. environmental, seasonal and physical changes), increasing false alarm rates over time and initiating the need for regular recalibration.

The latest evolution in video analytics is Advanced Video Pattern Analytics, which is based on pattern modelling algorithms, alerting on any change that has a pattern of a known object type such as a person or a vehicle. The technology focuses on recognising the objects in view and using information of the movement of the object to accurately classify it. Consider how humans recognise objects: we recognise an object based on its look, shape and movement. Advanced Video Pattern Detection works in a similar fashion.

Of the three types of video analytic technologies noted above, Advanced Video Pattern Analytics typically provides the lowest rate of false alarms, helping to sustain operator attention by highlighting information that is relevant and of interest.

Video analytics are becoming more intelligent, and feasible, as clients are looking for complete solutions as opposed to individual products. Complex integration of multiple systems, such as video management systems with other sub-systems like perimeter security systems demand rules that are based on integrated security management systems to manage increasing volumes of various different events. Ultimately, they provide better situational awareness and management of alarms.

In the never-ending battle between human and machine, Advanced Video Pattern Analytics provides a seamless work integration for the perfect balance of dependencies. Machine (technology) can only be good as the human’s actions on its alarms, and the human’s actions can only be as good as the machine’s quality of alarms.

Do you have this perfect balance in your control room?

Share this article:

Further reading: